An Intermediate Form of Behavioral Control in ‘Reactive’ Robots

The hypothesis of behavioral control as a form of traffic regulation is tested by means of simulation. Agents have to survive in an environment with a day/night rhythm by eating food and avoiding obstacles. This goal is achieved easily at day, but not at night due to interference of darkness with sensor readings. Two categories of agents are tested: reactive and control system agents. The control system is a neural network with inhibitory outputs to the agent’s behavioral layers. Therefore, the system is referred to as a traffic regulator. The neural network is optimized by an evolutionary algorithm. Results showed that control agents are able to adapt to the day/night rhythm by developing a day/night rhythm. These results are in favor of the traffic regulator hypothesis.