Autonomous robot path planning and obstacle avoidance in a dynamic environment

Abstract

An autonomous mobile robot can traverse an environment without human intervention. It uses path planning and obstacle avoidance techniques to find an optimal path from its start to goal location, without colliding into obstacles along the way. Such classical techniques include Artificial Potential Field, Cellular Decomposition and Vector Field Histograms. On the other hand, AI techniques include A*, Fuzzy Logic, Rapidly-exploring Random Trees, Ant Colony Optimisation, Particle Swarm Optimisation and Neural Networks. In this project, we investigated path planning algorithms that can be used in (a) the absence of obstacles, (b) in an environment with static obstacles and (c) in an environment with dynamic obstacles, meaning that some or all obstacles might be moving. We used the Elegoo Smart Robot Car Kit V4.0, an Arduino Uno as an Input/Output Subsystem and a Raspberry Pi 3 B+ for the processing of the path planning and obstacle avoidance techniques. We also used an ultrasonic sensor to detect obstacles around the robot and a gyroscope to measure the robot’s orientation. We controlled the robot’s motion using differential driving techniques and Proportional Integral Derivative Control. Firstly, we investigated a Search-Based algorithm referred to as Hybrid A*. Hybrid A* first uses the same technique as the A* algorithm to find a path using its knowledge of the environment. Then, Hybrid A* simplifies the path for the robot’s movement constraints. Secondly, we investigated a Sampling-Based algorithm referred to as Rapidly-exploring Random Tree*. This algorithm creates a tree by randomly generating nodes around the free space of the environment, until it reaches the goal. We tested both algorithms in three different environments: (a) without obstacles (b) with static obstacles (c) with dynamic obstacles. During all tests, both algorithms successfully guided the robot around the static and dynamic obstacles without any collision. However, we observed that Hybrid A* algorithms always managed to find the shorter path in all tests, together with having shorter computational time taken to find a path and the overall time taken to guide the robot from the start to finish.