Figure 4

The experimental results of the UAV reinforcement learning and reversal learning. (a) The training phase of the UAV two visual cues learning. A red fist is used to simulate heat punishment when the UAV is situated in front of inverted-blue T (image 3 and 7). (b–e) The test phase of the UAV two visual cues learning. We test the color learning ability by making choice on upright-green T (b) and upright-blue T (c). We test the shape learning ability by making choice on upright-white T (d) and inverted-white T (e). A blue rectangle is used as the reference object which is stable in different images. (f) The experimental results of reversal learning. The learning process contains two times of reverse and the y-axis is the chosen behavior. The blue bars denote the reversal learning process and red bars denote the learning ability on single cue. The red crosses represent incorrect choice and the red check marks represent the correct choice.