Real-Time Planning for Multiple Autonomous Vehicles in Dynamic Uncertain Environments

Abstract

This paper introduces a market-based cooperative planning system for a team of autonomous vehicles operating in a dynamic environment. The system combines the flexibility of evolution-computation techniques with the distributed nature of market strategy to compute task and paths plans. Optimization is based on a team utility function which accounts for uncertainty in knowledge of the environment. Multiple vehicles will cooperate on the same task if doing so increases the predicted team utility value. The team utility function and the associated stochastic model that predicts future system states are described. The minimum required information exchange among the vehicles is identified. Simulation results, using the Boeing Company developed Open Experimental Platform, demonstrate the effectiveness of the planning.

Collision avoidance in biological systems using collision cones

“…Prescribed maneuvers can be less effective than real time calculation methods as they cannot be modified to address specific situations. The optimization methods generally combine a kinematic model and a set of cost metrics to solve for trajectories with the lowest cost [9,25]. The costs that can be minimized can be economic costs like fuel and money or they can be workload.…”

Time constrained randomized path planning using spatial networks

“…Groups such as Capozzi et al have also looked at semi-randomized methods which provide quasi-optimal solutions to the path planning problem using evolutionary programming [3]. Later, Pongpunwattana et al incorporated these ideas into overall mission planning and task management schemes which address an agent's state and timing constraints [4]. Previous work investigated classical convex optimization techniques to generate simple paths from a starting point to a goal location for agents with constrained velocity limits [5].…”

Obstacle avoidance controller generating attainable set-points for the navigation of Multi-Robot System

“…The first one considers the robots control entirely based on path planning methods, which involve the prior knowledge of the robots environment. The objective is to find the best path to all the robots in order to avoid all the obstacles and each other while minimizing a cost function [1], [2]. This first method requires a significant computational complexity, especially when the environment is highly dynamic.…”