Solving the Real-Time Motion Planning Problem for Non-Holonomic Robots With Collision Avoidance in Dynamic Scenes

Collision-free motion planning allows robots to perform tasks in complex environments, but the conservative bounding curves during collision detection will reduce the robot's dexterity. This paper presents a novel algorithm for real-time motion planning of non-holonomic robots in dynamic scenes. To achieve this, the robot and the obstacles are generally decomposed into a series of superquadric objects. The expanded closed-form Minkowski sums are used to construct the velocity obstacle to deal with the updating orientation of the robot. The algorithm is extended to collision avoidance for different numbers and types of obstacles, which share the same workspace with the robot. We implement our algorithm in various simulations and experiments, where the robot has to avoid collisions with obstacles in real-time. The results demonstrate the effectiveness and efficiency of the proposed algorithm in collision-free motion planning at real-time computation rates and the enhanced dexterity control in complex environments.

I. Introduction

Collision-free motion planning is a classical problem in robotics. For instance, in a human-robot cooperative system, the robot must stay free of pedestrians. This problem can be formulated as an optimization problem for computing the collision-free path for the robot in environments, which contain other static and/or moving obstacles. In this direction, the robot is commanded by the pre-computed control input in each cycle based on environmental information and its desired target pose. Instead of a fixed path, the collision-free motion planning algorithm would compute a path that progresses towards the target as well as avoids collisions with obstacles.