Innovative Production Machines and Systems Conference

We examine the use of RL to aid a mobile robot in the nuclear decommissioning task of characterisation (i.e. the exploration of unknown or uncertain hazardous environments to assess and map levels of radiation so that safe and cost effective decontamination can subsequently be applied). Many existing robot nuclear characterisation systems are tethered, single robot systems. To speed up mapping and increase system fault tolerance it might be more appropriate to use multiple robots in which wireless network technologies rather than tethers are employed (despite interference or noise caused by radiation). While a wireless system may allow increased system flexibility and reduce inter-robot space conflict issues, a wireless robot is constrained by the need to periodically return to a recharging station to recharge its battery. We employ RL to enable a mobile robot to learn efficient paths back to a recharging station through an environment.