Real-time path planning for autonomous vehicle based on teaching–learning-based optimization

Abstract

This paper presents an online path planning approach for an autonomous tracked vehicle in a cluttered environment based on teaching–learning-based optimization (TLBO), considering the path smoothness, and the potential collision with the surrounding obstacles. In order to plan an efficient path that allows the vehicle to be autonomously navigated in cluttered environments, the path planning problem is solved as a multi-objective optimization problem. First, the vehicle perception is fully achieved by means of inertial measurement unit (IMU), wheels odometry, and light detection and ranging (LiDAR). In order to compensate the sensors drift to achieve more reliable data and improve the localization estimation and corrections, data fusion between the outputs of wheels odometry, LiDAR, and IMU is made through extended Kalman filter (EKF). Then, TLBO is proposed and applied to determine the optimum online path, where the objectives are to find the shortest path to reach the target destination, and to maximize the path smoothness, while avoiding the surrounding obstacles, and taking into account the vehicle dynamic and algebraic constraints. To check the performance of the proposed TLBO algorithm, it is compared in simulation to genetic algorithm (GA), particle swarm optimization (PSO), and a hybrid GA–PSO algorithm. Finally, real-time experiments based on robot operating system (ROS) implementation are conducted to validate the effectiveness of the proposed path planning algorithm.

Bioinspired Perception and Navigation of Service Robots in Indoor Environments: A Review

“…Inertial navigation usually is applied for reconfigurable robots [ 99 , 100 , 101 , 102 ], and some approaches uses sensor fusion techniques [ 96 , 97 , 98 ]. Inertial navigation can be utilized for exploration [ 96 , 99 ], sweeping [ 100 ], and autonomous transport [ 97 , 101 ]. Other navigation implements RFID or UWB for localization for entertainment [ 158 ] and searching [ 159 ].…”

“…However, most papers do not consider dynamic environments, except [ 13 , 136 , 154 ], which tried to solve the problem with learning approaches. Some studies indicate a dynamic environment as future work [ 13 , 93 , 97 , 124 , 157 ]. However, the problem of moving objects or dynamic obstacles is still not solved.…”

“…Sabiha et al [ 97 ] used teaching–learning-based optimization for online path planning in a cluttered environment, considering the potential collision and path smoothness as a multiobjective optimization problem, as shown in Figure 13 . An IMU achieved the robot perception by collaborating with LiDAR and wheel odometry, then via data fusion with the EKF.…”

Advancements in Learning-Based Navigation Systems for Robotic Applications in MRO Hangar: Review

“…Autonomous learning in path planning has made significant progress in recent times, where technologies such as CNN and deep reinforcement learning have been increasingly adopted. Path planning entails a sequence of configurations based on robot types and environment models that enable robots to navigate from a starting point to a target location [ 72 ]. The environment can be mapped to represent geometric information about the environment and connectivity between different nodes or maples.…”

Incorporating Learned Depth Perception Into Monocular Visual Odometry to Improve Scale Recovery

“…A vehicle operating autonomously requires precise and reliable information regarding its position for a successful navigation [1], [2], [3], [4]. There are currently many popular methods of providing comparatively reliable positioning information, such as the Global Navigation Satellite System (GNSS), Visual Odometry, and the Inertial Navigation System (INS) [5], [6], [7], [8], [9], [10], [11].…”