Stochastic Neural Analog Reinforcement Calculator

The Stochastic Neural Analog Reinforcement Calculator (SNARC) is a neural network machine designed by Marvin Minsky. Prompted by a letter from Minsky, George Armitage Miller gathered the funding (a few thousand dollars) for the project from the Office of Naval Research of the U.S. Department of Defense in the summer of 1951 with the work to be carried out by Minsky, who was then a graduate student in mathematics at Princeton University. At the time, a physics graduate student at Princeton, Dean S. Edmonds, volunteered that he was good with electronics and therefore Minsky brought him onto the project. During undergraduate years, Minsky was inspired by the 1943 Warren McCulloch and Walter Pitts paper on artificial neurons, and decided to build such a machine. The learning was Skinnerian reinforcement learning, and Minsky talked with Skinner extensively during the development of the machine. They tested the machine on a copy of Shannon's maze, and found that it could learn to solve the maze. Unlike Shannon's maze, this machine did not control a physical robot, but simulated rats running in a maze. The simulation is displayed as an "arrangement of lights", and the circuit was reinforced each time the simulated rat reached the goal. The machine surprised its creators. "The rats actually interacted with one another. If one of them found a good path, the others would tend to follow it." The machine itself is a randomly connected network of approximately 40 Hebb synapses. These synapses each have a memory that holds the probability that signal comes in one input and another signal will come out of the output. There is a probability knob that goes from 0 to 1 that shows this probability of the signals propagating. If the probability signal gets through, a capacitor remembers this function and engages an electromagnetic clutch. At this point, the operator will press a button to give a reward to the machine. This activates a motor on a surplus Minneapolis-Honeywell C-1 gyroscopic autopilot from a B-24 bomber. The motor turns a chain that goes to all 40 synapse machines, checking if the clutch is engaged or not. As the capacitor can only "remember" for a certain amount of time, the chain only catches the most recent updates of the probabilities. Each neuron contained 6 vacuum tubes and a motor. The entire machine is "the size of a grand piano" and contained 300 vacuum tubes. The tubes failed regularly, but the machine would still work despite failures. This machine is considered one of the first pioneering attempts at the field of artificial intelligence. Minsky went on to be a founding member of MIT's Project MAC, which split to become the MIT Laboratory for Computer Science and the MIT Artificial Intelligence Lab, and is now the MIT Computer Science and Artificial Intelligence Laboratory. In 1985 Minsky became a founding member of the MIT Media Laboratory. According to Minsky, he loaned the machine to students in Dartmouth, and subsequently lost, except for a single neuron. A photo of Minsky's last neuron can be seen here. The photo shows 6 vacuum tubes, one of which is a Sylvania JAN-CHS-6H6GT/G/VT-90A.

Geometric hashing

In computer science, geometric hashing is a method for efficiently finding two-dimensional objects represented by discrete points that have undergone an affine transformation, though extensions exist to other object representations and transformations. In an off-line step, the objects are encoded by treating each pair of points as a geometric basis. The remaining points can be represented in an invariant fashion with respect to this basis using two parameters. For each point, its quantized transformed coordinates are stored in the hash table as a key, and indices of the basis points as a value. Then a new pair of basis points is selected, and the process is repeated. In the on-line (recognition) step, randomly selected pairs of data points are considered as candidate bases. For each candidate basis, the remaining data points are encoded according to the basis and possible correspondences from the object are found in the previously constructed table. The candidate basis is accepted if a sufficiently large number of the data points index a consistent object basis. Geometric hashing was originally suggested in computer vision for object recognition in 2D and 3D, but later was applied to different problems such as structural alignment of proteins. == Geometric hashing in computer vision == Geometric hashing is a method used for object recognition. Let’s say that we want to check if a model image can be seen in an input image. This can be accomplished with geometric hashing. The method could be used to recognize one of the multiple objects in a base, in this case the hash table should store not only the pose information but also the index of object model in the base. === Example === For simplicity, this example will not use too many point features and assume that their descriptors are given by their coordinates only (in practice local descriptors such as SIFT could be used for indexing). ==== Training Phase ==== Find the model's feature points. Assume that 5 feature points are found in the model image with the coordinates ( 12 , 17 ) ; {\displaystyle (12,17);} ( 45 , 13 ) ; {\displaystyle (45,13);} ( 40 , 46 ) ; {\displaystyle (40,46);} ( 20 , 35 ) ; {\displaystyle (20,35);} ( 35 , 25 ) {\displaystyle (35,25)} , see the picture. Introduce a basis to describe the locations of the feature points. For 2D space and similarity transformation the basis is defined by a pair of points. The point of origin is placed in the middle of the segment connecting the two points (P2, P4 in our example), the x ′ {\displaystyle x'} axis is directed towards one of them, the y ′ {\displaystyle y'} is orthogonal and goes through the origin. The scale is selected such that absolute value of x ′ {\displaystyle x'} for both basis points is 1. Describe feature locations with respect to that basis, i.e. compute the projections to the new coordinate axes. The coordinates should be discretised to make recognition robust to noise, we take the bin size 0.25. We thus get the coordinates ( − 0.75 , − 1.25 ) ; {\displaystyle (-0.75,-1.25);} ( 1.00 , 0.00 ) ; {\displaystyle (1.00,0.00);} ( − 0.50 , 1.25 ) ; {\displaystyle (-0.50,1.25);} ( − 1.00 , 0.00 ) ; {\displaystyle (-1.00,0.00);} ( 0.00 , 0.25 ) {\displaystyle (0.00,0.25)} Store the basis in a hash table indexed by the features (only transformed coordinates in this case). If there were more objects to match with, we should also store the object number along with the basis pair. Repeat the process for a different basis pair (Step 2). It is needed to handle occlusions. Ideally, all the non-colinear pairs should be enumerated. We provide the hash table after two iterations, the pair (P1, P3) is selected for the second one. Hash Table: Most hash tables cannot have identical keys mapped to different values. So in real life one won’t encode basis keys (1.0, 0.0) and (-1.0, 0.0) in a hash table. ==== Recognition Phase ==== Find interesting feature points in the input image. Choose an arbitrary basis. If there isn't a suitable arbitrary basis, then it is likely that the input image does not contain the target object. Describe coordinates of the feature points in the new basis. Quantize obtained coordinates as it was done before. Compare all the transformed point features in the input image with the hash table. If the point features are identical or similar, then increase the count for the corresponding basis (and the type of object, if any). For each basis such that the count exceeds a certain threshold, verify the hypothesis that it corresponds to an image basis chosen in Step 2. Transfer the image coordinate system to the model one (for the supposed object) and try to match them. If successful, the object is found. Otherwise, go back to Step 2. === Finding mirrored pattern === It seems that this method is only capable of handling scaling, translation, and rotation. However, the input image may contain the object in mirror transform. Therefore, geometric hashing should be able to find the object, too. There are two ways to detect mirrored objects. For the vector graph, make the left side positive, and the right side negative. Multiplying the x position by -1 will give the same result. Use 3 points for the basis. This allows detecting mirror images (or objects). Actually, using 3 points for the basis is another approach for geometric hashing. === Geometric hashing in higher-dimensions === Similar to the example above, hashing applies to higher-dimensional data. For three-dimensional data points, three points are also needed for the basis. The first two points define the x-axis, and the third point defines the y-axis (with the first point). The z-axis is perpendicular to the created axis using the right-hand rule. Notice that the order of the points affects the resulting basis

Kullback–Leibler Upper Confidence Bound

In multi-armed bandit problems, KL-UCB (for Kullback–Leibler Upper Confidence Bound) is a UCB-type algorithm that is asymptotically optimal, in the sense that its regret matches the problem-dependent Lai-Robbins lower bound. == Multi-armed bandit problem == The Multi-armed bandit problem is a sequential game where one player has to choose at each turn between K {\displaystyle K} actions (arms). Behind every arm a {\displaystyle a} there is an unknown distribution ν a {\displaystyle \nu _{a}} that lies in a set D {\displaystyle {\mathcal {D}}} known by the player (for example, D {\displaystyle {\mathcal {D}}} can be the set of Gaussian distributions or Bernoulli distributions). At each turn t {\displaystyle t} the player chooses (pulls) an arm a t {\displaystyle a_{t}} , he then gets an observation X t {\displaystyle X_{t}} of the distribution ν a t {\displaystyle \nu _{a_{t}}} . === Regret minimization === The goal is to minimize the regret at time T {\displaystyle T} that is defined as R T := ∑ a = 1 K Δ a E [ N a ( T ) ] {\displaystyle R_{T}:=\sum _{a=1}^{K}\Delta _{a}\mathbb {E} [N_{a}(T)]} where μ a := E [ ν a ] {\displaystyle \mu _{a}:=\mathbb {E} [\nu _{a}]} is the mean of arm a {\displaystyle a} μ ∗ := max a μ a {\displaystyle \mu ^{}:=\max _{a}\mu _{a}} is the highest mean Δ a := μ ∗ − μ a {\displaystyle \Delta _{a}:=\mu ^{}-\mu _{a}} N a ( t ) {\displaystyle N_{a}(t)} is the number of pulls of arm a {\displaystyle a} up to turn t {\displaystyle t} The player has to find an algorithm that chooses at each turn t {\displaystyle t} which arm to pull based on the previous actions and observations ( a s , X s ) s < t {\displaystyle (a_{s},X_{s})_{s μ } {\displaystyle {\mathcal {K}}_{inf}(\nu ,\mu ,{\mathcal {D}}):=\inf \left\{\mathrm {KL} (\nu ,{\tilde {\nu }})\ |\ {\tilde {\nu }}\in {\mathcal {D}},\ \mathbb {E} [{\tilde {\nu }}]>\mu \right\}} K L {\displaystyle \mathrm {KL} } is the Kullback–Leibler divergence ν ^ a ( t ) {\displaystyle {\hat {\nu }}_{a}(t)} is the empirical distribution of arm a {\displaystyle a} at turn t {\displaystyle t} δ t {\displaystyle \delta _{t}} is a well-chosen sequence of positive numbers, often equal to ln ⁡ t + c ln ⁡ ln ⁡ t {\displaystyle \ln t+c\ln \ln t} with c > 0 {\displaystyle c>0} . Then we choose the arm a t {\displaystyle a_{t}} with the highest index: a t := arg ⁡ max a U a ( t ) {\displaystyle a_{t}:=\arg \max _{a}U_{a}(t)} We note that the algorithm does not require knowledge of T {\displaystyle T} . === Example === In the special case of Gaussian distribution with fixed variance σ 2 {\displaystyle \sigma ^{2}} , we have: U a ( t ) = μ ^ a ( t ) + 2 σ 2 δ t N a ( t ) {\displaystyle U_{a}(t)={\hat {\mu }}_{a}(t)+{\sqrt {\frac {2\sigma ^{2}\delta _{t}}{N_{a}(t)}}}} with μ ^ a ( t ) {\displaystyle {\hat {\mu }}_{a}(t)} being the empirical mean of arm a {\displaystyle a} at turn t {\displaystyle t} . === Pseudocode === The player gets the set D for each arm i do: n[i] ← 1; nu[i] ← None; d ← ln(K) for t from 1 to K do: select arm t observe reward r n[t] ← n[t] + 1 nu[t] ← update empirical distribution for t from K+1 to T do: for each arm i do: index[i] ← compute_index(n[i], nu[i], D, d) select arm a with highest index[a] observe reward r n[a] ← n[a] + 1 nu[a] ← update empirical distribution d ← ln(t+1) == Theoretical results == In the multi-armed bandit problem we have the Lai–Robbins asymptotic lower bound on regret. The algorithm KL-UCB matches this lower bound for one-dimensional exponential families with δ t := ln ⁡ t + 3 ln ⁡ ln ⁡ t {\displaystyle \delta _{t}:=\ln t+3\ln \ln t} and for distributions bounded in [ 0 , 1 ] {\displaystyle [0,1]} with δ t := ln ⁡ t + ln ⁡ ln ⁡ t {\displaystyle \delta _{t}:=\ln t+\ln \ln t} . === Lai–Robbins lower bound === In 1985 Lai and Robbins proved an asymptotic, problem-dependent lower bound on regret. It states that for every consistent algorithm on the set D {\displaystyle {\mathcal {D}}} — that is, an algorithm for which, for every ( ν 1 , … , ν K ) ∈ D K {\displaystyle (\nu _{1},\dots ,\nu _{K})\in {\mathcal {D}}^{K}} , the regret R T {\displaystyle R_{T}} is subpolynomial (i.e. R T = o T → + ∞ ( T α ) {\displaystyle R_{T}=o_{T\to +\infty }(T^{\alpha })} for all α > 0 {\displaystyle \alpha >0} ) — we have: R T ≥ ( ∑ a : μ a < μ ∗ Δ a K inf ( ν a , μ ∗ , D ) ) ln ⁡ T + o T → + ∞ ( ln ⁡ T ) . {\displaystyle R_{T}\geq \left(\sum _{a:\mu _{a}<\mu ^{}}{\frac {\Delta _{a}}{{\mathcal {K}}_{\inf }(\nu _{a},\mu ^{},{\mathcal {D}})}}\right)\ln T+o_{T\to +\infty }(\ln T).} This bound is asymptotic (as T → + ∞ {\displaystyle T\to +\infty } ) and gives a first-order lower bound of order ln ⁡ T {\displaystyle \ln T} with the optimal constant in front of it. === Regret bound for KL-UCB === The algorithm matches the Lai–Robbins lower bound for one-dimensional exponential-family distributions and for distributions bounded in [ 0 , 1 ] {\displaystyle [0,1]} . ==== One-dimensional exponential family ==== For D {\displaystyle {\mathcal {D}}} being the set of one-dimensional exponential families, with δ t := ln ⁡ t + 3 ln ⁡ ln ⁡ t {\displaystyle \delta _{t}:=\ln t+3\ln \ln t} we have the following upper bound on the regret of KL-UCB: R T ≤ ( ∑ a : μ a < μ ∗ Δ a K inf ( ν a , μ ∗ , D ) ) ln ⁡ T + O T ( ln ⁡ T ) . {\displaystyle R_{T}\leq \left(\sum _{a:\mu _{a}<\mu ^{}}{\frac {\Delta _{a}}{{\mathcal {K}}_{\inf }(\nu _{a},\mu ^{},{\mathcal {D}})}}\right)\ln T+O_{T}({\sqrt {\ln T}}).} ==== Bounded distributions in [0,1] ==== For D = P ( [ 0 , 1 ] ) {\displaystyle {\mathcal {D}}={\mathcal {P}}([0,1])} (the set of distributions supported on [ 0 , 1 ] {\displaystyle [0,1]} ), and for δ t := ln ⁡ t + ln ⁡ ln ⁡ t {\displaystyle \delta _{t}:=\ln t+\ln \ln t} , we have the following upper bound on the regret of KL-UCB: R T ≤ ( ∑ a : μ a < μ ∗ Δ a K inf ( ν a , μ ∗ , D ) ) ln ⁡ T + O T ( ( ln ⁡ T ) 4 / 5 ln ⁡ ln ⁡ T ) . {\displaystyle R_{T}\leq \left(\sum _{a:\mu _{a}<\mu ^{}}{\frac {\Delta _{a}}{{\mathcal {K}}_{\inf }(\nu _{a},\mu ^{},{\mathcal {D}})}}\right)\ln T+O_{T}{\big (}(\ln T)^{4/5}\ln \ln T{\big )}.} === Runtime === For D = P ( [ 0 , 1 ] ) {\displaystyle {\mathcal {D}}={\mathcal {P}}([0,1])} , the runtime needed per step and for an arm k {\displaystyle k} with n {\displaystyle n} observations is O ( n ( ln ⁡ n ) 2 ) {\displaystyle {\mathcal {O}}{\big (}n(\ln n)^{2}{\big )}} . This is higher than that of other optimal algorithms, such as NPTS with O ( n ) {\displaystyle {\mathcal {O}}(n)} . MED with O ( n ln ⁡ n ) {\displaystyle {\mathcal {O}}(n\ln n)} . and IMED with O ( n ln ⁡ n ) {\displaystyle {\mathcal {O}}(n\ln n)} . The high runtime of KL-UCB is due to a two-level optimisation: for each arm and candidate mean μ {\displaystyle \mu } , the algorithm evaluates K inf ( ν ^ a ( t ) , μ , D ) {\displaystyle {\mathcal {K}}_{\inf }({\hat {\nu }}_{a}(t),\mu ,{\mathcal {D}})} and then maximises μ {\displaystyle \mu } subject to N a ( t ) K inf ( ν ^ a ( t ) , μ , D ) ≤ δ t {\displaystyle N_{a}(t)\,{\mathcal {K}}_{\inf }({\hat {\nu }}_{a}(t),\mu ,{\mathcal {D}})\leq \delta _{t}} . For distributions bounded in [ 0 , 1 ] {\displaystyle [0,1]} the inner problem has no closed form and must be solved numerically, which increases the per-step cost.

Webometrics

The science of webometrics (also referred to as cybermetrics) aims to quantify the World Wide Web to get knowledge about the number and types of hyperlinks, the structure of the World Wide Web, and using patterns. According to Björneborn and Ingwersen, the definition of webometrics is "the study of the quantitative aspects of the construction and use of information resources, structures and technologies on the Web drawing on bibliometric and informetric approaches." The term webometrics was coined by Almind and Ingwersen (1997). A second definition of webometrics has also been introduced, "the study of web-based content with primarily quantitative methods for social science research goals using techniques that are not specific to one field of study", which emphasizes the development of applied methods for use in the wider social sciences. The purpose of this alternative definition was to help publicize appropriate methods outside the information-science discipline rather than to replace the original definition within information science. Similar scientific fields are: bibliometrics, informetrics, scientometrics, virtual ethnography, and web mining. One relatively straightforward measure is the "web impact factor" (WIF) introduced by Ingwersen (1998). The WIF measure may be defined as the number of web pages in a web site receiving links from other web sites, divided by the number of web pages published in the site that are accessible to the crawler. However, the use of WIF has been disregarded due to the mathematical artifacts derived from power law distributions of these variables. Other similar indicators using size of the institution instead of number of webpages have been proved more useful.

Skyline operator

The skyline operator is the subject of an optimization problem and computes the Pareto optimum on tuples with multiple dimensions. This operator is an extension to SQL proposed by Börzsönyi et al. to filter results from a database to keep only those objects that are not dominated by any other point on all dimensions. The name skyline comes from the view on Manhattan from the Hudson River, where those buildings can be seen that are not hidden by any other. A building is visible if it is not dominated by a building that is taller or closer to the river (two dimensions, distance to the river minimized, height maximized). Another application of the skyline operator involves selecting a hotel for a holiday. The user wants the hotel to be both cheap and close to the beach. However, hotels that are close to the beach may also be expensive. In this case, the skyline operator would only present those hotels that are not worse than any other hotel in both price and distance to the beach. == Formal specification == The skyline operator returns tuples that are not dominated by any other tuple. A tuple dominates another if it is at least as good in all dimensions and better in at least one dimension. Formally, we can think of each tuple as a vector p , q ∈ R n {\displaystyle p,q\in \mathbb {R} ^{n}} . p {\displaystyle p} dominates q {\displaystyle q} (written: p ≻ q {\displaystyle p\succ q} ) if p {\displaystyle p} is at least as good as q {\displaystyle q} in every dimension, and superior in at least one: p ≻ q ⇔ ∀ i ∈ [ n ] . p [ i ] ⪰ q [ i ] ∧ ∃ j ∈ [ n ] . p [ j ] ≻ q [ j ] . {\displaystyle p\succ q\Leftrightarrow \forall i\in [n].p[i]\succeq q[i]\wedge \exists j\in [n].p[j]\succ q[j].} Dominance ( p ≻ q {\displaystyle p\succ q} ) can be defined as any strict partial ordering, for example greater (with ≻:=> {\displaystyle \succ :=>} and ⪰:=≥ {\displaystyle \succeq :=\geq } ) or less (with ≻:=< {\displaystyle \succ :=<} and ⪰:=≤ {\displaystyle \succeq :=\leq } ). Assuming two dimensions and defining dominance in both dimensions as greater, we can compute the skyline in SQL-92 as follows: == Proposed syntax == As an extension to SQL, Börzsönyi et al. proposed the following syntax for the skyline operator: where d1, ... dm denote the dimensions of the skyline and MIN, MAX and DIFF specify whether the value in that dimension should be minimised, maximised or simply be different. Without an SQL extension, the SQL query requires an antijoin with not exists: == Implementation == The skyline operator can be implemented directly in SQL using current SQL constructs, but this has been shown to be very slow in disk-based database systems. Other algorithms have been proposed that make use of divide and conquer, indices, MapReduce and general-purpose computing on graphics cards. Skyline queries on data streams (i.e. continuous skyline queries) have been studied in the context of parallel query processing on multicores, owing to their wide diffusion in real-time decision making problems and data streaming analytics. Exasol features a native implementation.

INDIAai

INDIAai is a web portal launched by the Government of India on 07 March 2024 for artificial intelligence-related developments in India. It is known as the National AI Portal of India, which was jointly started by the Ministry of Electronics and Information Technology (MeitY), the National e-Governance Division (NeGD) and the National Association of Software and Service Companies (NASSCOM) with support from the Department of School Education and Literacy (DoSE&L) and Ministry of Human Resource Development. == History == The portal was launched on 30 May 2020, by Ravi Shankar Prasad, the Union Minister for Electronics and IT, Law and Justice and Communications, on the first anniversary of the second tenure of Prime Minister Narendra Modi-led government. A national program for the youth, 'Responsible AI for Youth', was also launched on the same day. As of 2022, the website was visited by more than 4.5 lakh users with 1.2 million page views. It has 1151 articles on artificial intelligence, 701 news stories, 98 reports, 95 case studies and 213 videos on its portal. It maintains a database on AI ecosystem of India featuring 121 government initiatives and 281 startups. In May 2022, INDIAai released a book titled 'AI for Everyone' that covers the basics of AI. Cabinet chaired by the Prime Minister Narendra Modi has approved the comprehensive national-level IndiaAI mission with a budget outlay of Rs.10,371.92 crore. The Mission will be implemented by ‘IndiaAI’ Independent Business Division (IBD) under Digital India Corporation (DIC). == Objective and features == It aims to function as a one-stop portal for all AI-related development in India. The platform publishes resources such as articles, news, interviews, and investment funding news and events for AI startups, AI companies, and educational firms related to artificial intelligence in India. It also distributes documents, case studies, and research reports. Additionally, the platform provides education and employment opportunities related to AI. It offers AI courses, both free and paid.

Kinodynamic planning

In robotics and motion planning, kinodynamic planning is a class of problems for which velocity, acceleration, and force/torque bounds must be satisfied, together with kinematic constraints such as avoiding obstacles. The term was coined by Bruce Donald, Pat Xavier, John Canny, and John Reif. Donald et al. developed the first polynomial-time approximation schemes (PTAS) for the problem. By providing a provably polynomial-time ε-approximation algorithm, they resolved a long-standing open problem in optimal control. Their first paper considered time-optimal control ("fastest path") of a point mass under Newtonian dynamics, amidst polygonal (2D) or polyhedral (3D) obstacles, subject to state bounds on position, velocity, and acceleration. Later they extended the technique to many other cases, for example, to 3D open-chain kinematic robots under full Lagrangian dynamics. == Modern approaches == Since the foundational theoretical work of the 1990s, the field has evolved significantly with new algorithmic approaches that address the computational and practical limitations of early methods. === Sampling-based methods === Many practical heuristic algorithms based on stochastic optimization and iterative sampling have been developed by a wide range of authors to address the kinodynamic planning problem. Popular approaches include extensions of RRT algorithms such as RRT for kinodynamic systems, and sampling-based methods like Model Predictive Path Integral (MPPI) control. These stochastic techniques have been shown to work well in practice and can handle complex, high-dimensional state spaces more efficiently than deterministic methods. However, all motion planning methods are subject to the PSPACE-hardnesss of classical motion planning even without dynamics, which means (assuming the usual structural complexity conjectures) they all can be worst-case exponential-time in the state-space dimension (the number of degrees of freedom). On the other hand, the deterministic methods have provable guarantees of completeness, accuracy, and complexity (for fixed dimension, they are polynomial-time not only in the geometric complexity, but also in ( 1 / ε ) {\displaystyle (1/\varepsilon )} , the closeness of the desired approximation), whereas most of the recent heuristic/stochastic methods sacrifice at least one of these criteria. === Mixed-integer optimization approaches === Recent advances in mixed-integer programming have enabled new deterministic approaches to kinodynamic planning. These methods formulate the planning problem as an optimization task that simultaneously determines the spatial path and control sequence while respecting all kinodynamic constraints. By using techniques such as McCormick envelopes to handle bilinear constraints, these approaches can provide globally optimal solutions with mathematical guarantees while achieving significant computational speedups over traditional methods. === Genetic algorithm approaches === Genetic algorithms have also been adapted for kinodynamic planning, particularly for gradient-free optimization in challenging terrain. These methods use evolutionary computation to optimize trajectories over receding horizons, with specialized mutation operators that ensure vehicle controls remain within operational limits. This approach is particularly useful when dealing with non-differentiable cost functions or when gradient information is unavailable or unreliable. === Three-dimensional terrain planning === The foundational theoretical work of the 1990s was extended to higher degrees of freedom, and even to n {\displaystyle n} -link, 3D open-chain kinematic robots under full Lagrangian dynamics. However, many of the subsequent heuristic techniques (typically employing stochastic optimization) were confined to planar environments. More recent kinodynamic planning has extended beyond these planar environments to handle complex 3D terrains represented as simplicial complexes or triangular meshes. This advancement is particularly important for applications such as autonomous vehicle navigation in off-road environments, where elevation changes and terrain geometry significantly impact vehicle dynamics. These methods must account for pitch angles, surface curvature, and the coupling between terrain geometry and vehicle kinodynamic constraints. == Performance and guarantees == The landscape of performance guarantees in kinodynamic planning has evolved considerably. While early heuristic methods could not guarantee optimality, recent mixed-integer approaches have demonstrated the ability to find globally optimal solutions with proven constraint satisfaction. Experimental comparisons have shown that modern optimization-based planners can achieve execution times several orders of magnitude faster than sampling-based methods while maintaining strict adherence to kinodynamic constraints. However, the choice of method often depends on the specific application requirements. Sampling-based methods remain valuable for their ability to quickly find feasible solutions in high-dimensional spaces and their robustness to modeling uncertainties. Optimization-based methods excel when optimality guarantees and constraint compliance are critical, particularly in safety-critical applications. == Applications == Kinodynamic planning finds applications across numerous domains including: Autonomous vehicles: Path planning for cars, trucks, and other ground vehicles that must respect acceleration, steering, and velocity limits Aerial robotics: Trajectory planning for quadrotors and other unmanned aerial vehicles with dynamic constraints Manipulation: Planning for robotic arms where joint velocities, accelerations, and torques are limited Legged locomotion: Footstep and trajectory planning for walking and running robots Space robotics: Planning under thrust and fuel constraints for spacecraft and rovers