Motion planning

About: Motion planning is a research topic. Over the lifetime, 32846 publications have been published within this topic receiving 553548 citations.

Coordinating the motions of multiple robots with specified trajectories

Abstract: Coordinating the motions of multiple robots operating in a shared workspace without collisions is an important capability. We address the task of coordinating the motions of multiple robots when their trajectories (defined by both the path and velocity along the path) are specified. This problem of collision-free trajectory coordination arises in welding and painting workcells in the automotive industry. We identify sufficient and necessary conditions for collision-free coordination of the robots when only the robot start times can be varied, and define corresponding optimization problems. We develop mixed integer programming formulations of these problems to automatically generate minimum time solutions. This method is applicable to both mobile robots and articulated arms, and places no restrictions on the number of degrees of freedom of the robots. The primary advantage of this method is its ability to coordinate the motions of several robots, with as many as 20 robots being considered. We show that, even when the robot trajectories are specified, minimum time coordination of multiple robots is NP-hard.

Mobile robot navigation and target tracking system

Abstract: This paper presents the framework for the navigation and target tracking system for a mobile robot. Navigation and target tracking are to be performed using a Microsoft Xbox Kinect sensor which provides RGB color and 3D depth imaging data to an x86 based computer onboard the robot running Ubuntu Linux. A fuzzy logic controller to be implemented on the computer is considered for control of the robot in obstacle avoidance and target following. Data collected by the computer is to be sent to a server for processing with learning-based systems utilizing neural networks for pattern recognition, object tracking, long-term path planning and process improvement. An eventual goal of this work is to create a multi-agent robot system that is able to work autonomously in an outdoor environment.

BayesSim: Adaptive Domain Randomization Via Probabilistic Inference for Robotics Simulators

Abstract: We introduce BayesSim, a framework for robotics simulations allowing a full Bayesian treatment for the parameters of the simulator. As simulators become more sophisticated and able to represent the dynamics more accurately, fundamental problems in robotics such as motion planning and perception can be solved in simulation and solutions transferred to the physical robot. However, even the most complex simulator might still not be able to represent reality in all its details either due to inaccurate parametrization or simplistic assumptions in the dynamic models. BayesSim provides a principled framework to reason about the uncertainty of simulation parameters. Given a black box simulator (or generative model) that outputs trajectories of state and action pairs from unknown simulation parameters, followed by trajectories obtained with a physical robot, we develop a likelihood-free inference method that computes the posterior distribution of simulation parameters. This posterior can then be used in problems where Sim2Real is critical, for example in policy search. We compare the performance of BayesSim in obtaining accurate posteriors in a number of classical control and robotics problems. Results show that the posterior computed from BayesSim can be used for domain randomization outperforming alternative methods that randomize based on uniform priors.

Languages, behaviors, hybrid architectures, and motion control

Abstract: In this paper we put forward a framework that integrates features of reactive planning models with modern control-theory-based approaches to motion control of robots. We introduce a motion description language, MDLe, inspired by Roger Brockett’s MDL, that provides a formal basis for robot programming using behaviors, and at the same time permits incorporation of kinematic and dynamic models of robots given in the form of differential equations. In particular, behaviors for robots are formalized in terms of kinetic state machines, a motion description language, and the interaction of the kinetic state machine with realtime information from (limited range) sensors. This formalization allows us to create a mathematical basis for the study of such systems, including techniques for integrating sets of behaviors. In addition we suggest cost functions for comparing both atomic and compound behaviors in various environments. We demonstrate the use of MDLe in the area of motion planning for nonholonomic robots. Such models impose limitations on stabilizability via smooth feedback; piecing together open-loop and closed-loop trajectories becomes essential in these circumstances, and MDLe enables one to describe such piecing together in a systematic manner. A reactive planner using the formalism of this discussion is described. We demonstrate obstacle avoidance with limited range sensors as a test of this planner.