Siemens Aktiengesellschaft

Thilo Grundmann

Raoul Daniel Zöllner

A decision mechanism is configured to decide on at least one prospective action of a robot from set of actions by: computing a prior probabilistic representation of a prior environment state; updating the prior probabilistic representation with targets of a new observation on reducing at least one uncertainty in a posterior probabilistic representation of a posterior environment state to be reached after an appliance of the at least one prospective action, the posterior probabilistic representation resulting from the updating; determining an information gain between the prior probabilistic representation and the posterior probabilistic representation by use of at least one information theoretic measure; evaluating the at least one prospective action by adding costs of executing the at least one prospective action to the information gain. Furthermore, an improved action planning for robots is provided and can be implemented in various robots investigating scenes for their actions.

Growing community of inventors

Munich, Germany

Thilo Grundmann