Fast and Accurate Collision Probability Estimation for Autonomous Vehicles using Adaptive Sigma-Point Sampling

A novel algorithm is presented for the estimation of collision probabilities between dynamic objects with uncertain trajectories, where the trajectories are given as a sequence of poses with Gaussian distributions. We propose an adaptive sigma-point sampling scheme, which ultimately produces a fast, simple algorithm capable of estimating the collision probability with a median error of 3.5%, and a median runtime of 0.21ms, when measured on an Intel Xeon Gold 6226R Processor. Importantly, the algorithm explicitly accounts for the collision probability's temporal dependence, which is often neglected in prior work and otherwise leads to an overestimation of the collision probability. Finally, the method is tested on a diverse set of relevant real-world scenarios, consisting of 400 6-second snippets of autonomous vehicle logs, where the accuracy and latency is rigorously evaluated.

Subgoal Planning Algorithm for Autonomous Vehicle Guidance

Trained humans exhibit highly agile spatial skills, enabling them to operate vehicles with complex dynamics in demanding tasks and conditions. Prior work shows that humans achieve this performance by using strategies such as satisficing, learning hierarchical task structure, and using a library of motion primitive elements. A key aspect of efficient and versatile solutions is the decomposition of a task into a sequence of smaller tasks, represented by subgoals. The present work uses properties of constrained optimal control theory to define conditions that specify candidate subgoal states and enable this decomposition. The proposed subgoal algorithm uses graph search to determine a subgoal sequence that links a series of unconstrained motion guidance elements into a constrained solution trajectory. In simulation experiments, the subgoal guidance algorithm generates paths with higher performance and less computation time than an RRT* benchmark. Examples illustrate the robustness and versatility of this approach in multiple environment types.

First-Person Perceptual Guidance Behavior Decomposition using Active Constraint Classification

Humans exhibit a wide range of adaptive and robust dynamic motion behavior that is yet unmatched by autonomous control systems. These capabilities are essential for real-time behavior generation in cluttered environments. Recent work suggests that human capabilities rely on task structure learning and embedded or ecological cognition in the form of perceptual guidance. This paper describes the experimental investigation of the functional elements of human motion guidance, focusing on the control and perceptual mechanisms. The motion, control, and perceptual data from first-person guidance experiments is decomposed into elemental segments based on invariants. These elements are then analyzed to determine their functional characteristics. The resulting model explains the structure of the agent-environment interaction and provides lawful descriptions of specific perceptual guidance and control mechanisms.