Optimal Robot Recharging Strategies For Time Discounted Labour

Energy is defined as the potential to perform work: every system that does some work must possess the required energy in advance. An interesting class of systems, including animals and recharging robots, has to actively choose when to obtain energy and when to dissipate energy in work. If working and collecting energy are mutually exclusive, as is common in many animal and robot scenarios, the system faces an essential two-phase action selection problem: (i) how much energy should be accumulated before starting work; (ii) at what remaining energy level should the agent switch back to feeding/recharging? This paper presents an abstract general model of a energy-managing agent that does time-discounted work. Analyzing the model, we find solutions to both questions that optimise’s the value of the work done. This result is validated empirically by simulated robot experiments that agree closely with the model.