For autonomous mobile robots capable of self-docking to a battery charging station, the standard solution for battery management is to set up a threshold of battery level chosen so that any risk of battery depletion is avoided. We introduce the idea that battery management of mobile robots should not consist of simple avoidance of battery depletion: in certain situations, taking risk of battery depletion is justified in regard to the level of mission accomplishment that can be attained.

Based on this consideration, we proposed an autonomous battery management for mobile robots based on risk and gain assessment, gain being defined as the level of mission accomplishment reached by the robot measured in percentage. Using the proposed risk/gain battery management method, the robot decides when to redirect itself to the station in two steps:

- It determines if, in the time frame it can estimate future values of risk and gain, risk is always balanced by expected gain. What level of gain is required for balancing a given level of risk depends on a configurable risk-taking parameter. To estimate the risk of battery depletion, a method based on the use of probability density functions is proposed.

- In the window of time corresponding to accepted levels of risk, a time corresponding to a large risk/gain gap is selected as suitable for redirection to the station, the term gap referring to the simple substraction of the normalized gain with the risk of battery depletion.

Assessed risk can be used along with measure of gain to calculate, depending on a risk-taking parameter, the best time to redirect the robot to the docking station. Contrary to the use of a fixed battery threshold, the proposed approach gives flexibility to the robot to adapt to its current situation and to insure that higher risk will correspond to higher level of mission accomplishment. An implementation of the system allowed a room cleaning robot to autonomously decide to redirect itself to the station before a sharp increase of risk not corresponding to a significant extension of operational time.