They predominantly use ultrasonic echoes to perceive their surroundings and fly through dense forests in complete darkness with ease. The question is then, “given the information about multiple sensed obstacles and the target location, how do we combine them to select a good path?”Įcholocating bats are excellent examples of creatures that possess such a capability. Philosophically, obstacles should not determine the direction in which a creature should move, rather they should simply indicate where the creature should not go. Although reflexive behaviors are well-suited to a creature traveling quickly through an unknown or changing sparse environment, even mildly cluttered environments can produce inappropriate movements.
In contrast, reflexive algorithms simply steer the creature away from obstacles upon detection with very little latency ( Milde et al., 2017). These are appropriate when a tremendous amount of relevant knowledge about the world is available and optimal paths are desired. Path planning algorithms calculate routes between starting and goal points, requiring extensive knowledge of the environment and accurate localization. Clearly, there is an expansive world of algorithms lying between these two extremes. In the world of robotics, there have historically been two extreme philosophical starting points in the approach to solving this problem: rigorous path planning assuming accurate and extensive sensing ( Latombe, 2012) and fast reflexive behaviors based on minimal and unsophisticated sensing ( Braitenberg, 1986). In addition, animals are often observed to orient their heads in different directions to gather sensory information needed for obstacle avoidance.
Animals are often able to detect obstacles using different types of sensors to quickly decide on the motions to avoid them. Traveling through an environment toward a goal without colliding with obstacles is one of many essential abilities for animals to survive. We further propose a spiking neural model using spike-timing representations, a spike-latency memory, and a “race-to-first-spike” WTA circuit. A mobile robot was used to test the proposed model navigating through a cluttered environment using a narrow field-of-view sonar system. To avoid excessive scanning of the environment, an attentional system is proposed to control the directions of sonar pings for efficient, task-driven, sensory data collection. A two-stage winner-take-all (WTA) mechanism is used to select the final steering action. Taking into account the limited field-of-view of practical sonar systems and vehicle kinematics, we propose a neural model for obstacle avoidance that maps the 2-D sensory space into a 1-D motor space and evaluates motor actions while combining obstacles and goal information.
The rapid control of a sonar-guided vehicle to pursue a goal while avoiding obstacles has been a persistent research topic for decades.
0 kommentar(er)
