Page 28 - Read Online

Li et al. Intell Robot 2021;1(1):58-83 I http://dx.doi.org/10.20517/ir.2021.08 Page 78

• Most studies on the navigation and control of robots fail not to consider the teleoperation and telepresence
issues. It is assumed that the robot works without human interactions. New approaches to telerobotic
operations and human-robot interactions would be developed based on biologically inspired intelligence
tooutperformexistingtechnologies. Thefuturedevelopedalgorithmswillnotdirectlymimicanybiological
systems. The infusion of “human-like” and biological intelligence into robotic systems is the crux of future
research.

6. CONCLUSION
Biologically inspired intelligence has been explored and studied for decades in the field of robotics. The re-
searchers have been trying to replicate or transfer the biological intelligence to robotic systems for empow-
ering the robots stability, adaptability, and cooperativeness. This paper provides a comprehensive survey of
the research on bio-inspired neurodynamics models and their applications to path planning and control of
autonomous robots. Among all bio-inspired neurodynamics models, shunting models, additive models, and
gated dipole models were further elaborated. As for path planning, a bio-inspired neural network was elabo-
rated for the dynamic collision-free path generation for many robotic systems. There are several key points
are worth to highlight about bio-inspired neurodynamics models to real-time collision-free path planning.

• The fundamental concept of the neurodynamics-based path planning approach is to develop a one-to-one
correspondence neural network, which is called the bio-inspired neural network, to represent the work
environment. The neural activity is a continuous signal with both upper and lower bounds.
• Thebio-inspiredneuralnetworkisabletoguidetherobottoavoidthelocalminimapointsandthedeadlock
situations. The target globally influences the whole work space through neural activity propagation to all
directions in the same manners.
• The bio-inspired neural network is able to generate the path without explicitly searching over the free work
spaceorthecollisionpaths,withoutexplicitlyoptimizinganyglobalcostfunctions,withoutanypriorknowl-
edge of the dynamic environment, and without any learning procedures.
• The bio-inspired neural network is able to perform properly in an arbitrarily dynamic environment, even
with sudden environmental changes, such as suddenly adding or removing obstacles or targets. The obsta-
cles have only local effects to push the robot to avoid collisions.

As for the bio-inspired robot control, several key points are worth to note:

• The neural activity is bounded between the [− , ] region with different inputs, which is the fundamental
concept of the bio-inspired backstepping control.
• The bio-inspired backstepping control provides a smooth velocity curve, which is crucial to ensure the
control effectiveness and efficiency
• The speed jump in conventional backstepping control design is eliminated by replacing the tracking error
term with shunting model, this modification allows a wider application of the bio-inspired backstepping
control in robotics.
• The excellent feasibility of the bio-inspired backstepping control allows it compatible with many other con-
trol strategies to form new hybrid control strategies for robots working in various working environments.

The current challenges and future works are the development of original and innovative new intelligent navi-
gation, cooperation, and communication strategies, algorithms and technologies with consideration of uncer-
tainties, disturbance and noise issues, communication issues, and human-robot interaction issues for robots
in changing complex situations.