Page 19 - Read Online

Page 69 Li et al. Intell Robot 2021;1(1):58-83 I http://dx.doi.org/10.20517/ir.2021.08

Figure 8. An example of path planning in 3-D underwater environments [25] .

titioned into small portions centered with moving AUV, and the bio-inspired neural network only deals with
this small range, the path can be calculated relatively fast. However, the dynamic bio-inspired neural network
misses the best route in certain circumstances, which could waste the power of the AUV.

The environmental disturbances of the underwater area, such as currents, create inevitable influences on the
AUV path planning. A current effect-eliminated bio-inspired neural network was proposed to guide the AUV
navigation considering the effect of currents [70] . A current correcting component was incorporated with the
bio-inspired neural network to generate the paths. Each neuron in the network, the velocity and direction of
robots are corrected for eliminating the current effect. Thus, the generated UUV path is robust and efficient.

The real-world ocean environment is complex and unknown. The onboard robot sensors were used for robot
navigationwithalimiteddetectionrange. Theultrasonicsensorwasusedtointerpretthesonardataandupdate
the map based on the dempster’s inference rule [71] . A potential field bio-inspired neural network (PBNN) was
proposed to generate a safe path in underwater environments [72,73] . The planned path keeps a safe distance to
the obstacles, which could avoid the collisions for the underwater robot navigation.

Multi-AUV systems cooperation has received lots of interest due to the fact that groups of AUVs can work
more efficiently and effectively compared with a single AUV. The main task of AUVs cooperation is to assign
several targets to a team of AUVs and avoid obstacles autonomously in underwater environments. Due to
the similarity of multi-tasks assignment and self-organizing map (SOM) neural networks, many researchers
have been applied the SOM approach to solve task assignment problems of multi-robotic systems [74–76] and
multi-AUV systems [77,78] . However, the SOM-based methods require an ideal 2-D work environment without
obstacles. An integrated biologically inspired SOM (BISOM) method was proposed to deal with collision-free
and multi-AUV task assignment problems [79] . After integrated the bio-inspired neural network method, the
AUV is able to avoid obstacles and speed jumps. The ocean currents could influence the AUV navigation in
the underwater environment. A velocity synthesis algorithm was integrated with the BISOM approach for
optimizing the individual robot path in a dynamic environment considering the ocean current [80] .

The BISOM method is able to generate the shortest path for the multi-robotic systems in most situations.
However, the update rule of the BISOM method ignores the effect of obstacles. Therefore, although the win-
ner AUV is the shortest distance from the target, the obstacles could increase the movement of the winner
robot. A novel biologically inspired map algorithm was proposed by Zhu et al. [81] for changing the update