Performance of two homing strategies in environments with differently distributed obstacles

Two computationally simple forms of dead reckoning were implemented in a physical robot (Khepera). The precision of homing was compared between both strategies for various environments in which obstacles were distributed in four different ways: regular, at random, clumped, and with walls. Counter intuitively, the computational more complex strategy did not perform better in general. Both strategies performed best in a regular environment. In a regular environment the simpler strategy ('counter turning') was even more precise than the more complex strategy ('retrace your trail'). On the other hand retrace your trail performed better than counter turning in the clumped environment.