Reducing operational cost of an autonomous robot using combined data-driven and navigational algorithms
Abstract
Autonomous mobile robots have become an important presence in various fields of work, but are restricted by short operation times due to the limitations of their onboard energy sources. This factor bottlenecks applications of mobile robots in many situations. In this thesis, a data-driven power consumption model is developed based on the motion data of the robot. A path planning algorithm combining the rapid-exploring random tree and artificial potential field was developed for navigation. From this, a combined framework utilizing model predictive control as the control strategy to optimize motion and conserve power is proposed. The algorithm is implemented in a popular research platform, TurtleBot3, for validation in a dynamic environment. These tests were conducted on a flat surface using five different obstacle configurations, with obstacles being hidden from initial detection. The experimentation results demonstrate the effectiveness of the proposed algorithm framework in reducing the required power and calculation time.