(Above: Honda Asimov Humanoid Robot)Often I am asked what is the main reason behind designing humanoid robots. Most people are scared that humanoid robots may be competitive to their biological counterparts. In this article I discuss the most basic advantages of developing and incorporating humanoid robots in near future. Some enabling technologies to develop humanoid robots include telepresence, series elastic actuators, passive dynamic control and machine vision.
1. So Why Humanoid Robots ?
Many laboratories today are exploring semi or full humanoid robot platforms to reverse engineer biomechanics and biomechanical control. The priniciple idea behind a humanoid robot is to mimic human body structure and motion (kinematics and dynamics) along with control aspects(neural control of movement). Human brain is difficult to understand, we know more about the structure of the universe then we know about our brain and its mysteries. Humanoid robotics maybe one way to reverse engineer neural pathways and understand biomechanical models of motion and cognition. Control schemes to balance a humanoid robot or provide dexterous manipulation results from detailed computational models which can shed inside on the neurobiological and cognitive abilities exhibited by human brain. Another characteristic about humanoid robots is that they are mobile and unlike wheeled bases they can climb stairs and negotiate complex terrain at variable speeds thus providing a highly mobile platform that gathers information (by deploying sensors) in "real world" and dynamically interacting with it. Main advantage of humanoid robot platform is that unlike the fixed base factory manipulators and other manipulator arms and wheeled robots that work in highly structured environments, a humanoid platform mimics its biological counterpart in its ability to explore unstructured terrain. Such abilities both in structure and control can provide indespensible manipulative, perceptive and communicative skills to do real time interaction with humans. A humanoid robot platform in the near future can provide assistance to humans in different scenarios. I am currently assisting research scientists at MIT to develop partial (upper torso) humanoid like structures with SEA based joints to lift wounded soldiers for safe battle field extraction. Other applications that I am working on include man machine technologies only possible with humanoid or partial humanoid like structures like class room interactions using telepresence robots and domestic mobile robots called Partner robots. Several famous researchers at MIT are pushing the limits in humanoid robot design by working on invidiual joints like ankles and knees which may someday be integrated into a full humanoid. Such units can also be used in conjunction with biological devices to provide an orthosis or prosthesis that is controlled under neural commands. While these applications will be feasible in the near future currently several upper body torso's have already been designed and tested for example: DOMO , OBERO , MDS
etc. These robots have steerable head with stereo binocular vision and arms and torso supported on wheeled bases. The control of bipedal walking is highly expensive (power) and is a possibility in near future with efficient algorithms.
2. Design of Humanoid Robots
2.1 Mechanical Design
The humanoid robots found today mimic biomechanics only at a very high level for example all the bone structures are modeled rigid in the mechanical design of the humanoid robot joints, while the biological systems are compliant. Researchers at MIT are working on safe and compliant actuators called Series Elastic Actuators (SEA) which provide safety (to the robot itself and to the people) and can mimic the biological properties of muscle and tendon to first order approximation. Another limitation of today's humanoid robot's lies in the uniarticular and bi-articular implementation of the SEA based muscles due to limitations in clutch technology.
2.2 Control Systems
Humanoid robots are made primarily to behave like humans, hence ordinary industrial control systems may not be sufficient to provide the bandwidth and the biomimetic characteristics necessary for controlling the electric or hydraulic actuators that make up muscles of the humanoid robot. Also traditional industrial control systems like PLC's cannot be used since they maybe too big or bulky for mobile robotic platforms like humanoid robots. The control of humanoid robot can be divided into two separate control problems namely the upper extremity and the lower extremity. The upper extremity control problem for humanoid robots has been successfully solved with several robots like Honda ASIMOV capable of advanced dexterous behaviors like holding human hand, following the human hand with visual servoing etc. The extremely challenging part of humanoid robot control system is the bipedal walking, running, stair climbing with minimum energy. Researchers at MIT CSAIL, MIT Media Lab, MIT Aero Astro etc have been making tremendous progress in modeling the lower extremity as inverted pendula and solving the contact dynamics in walking,running and developing associated muscle models for detailed biomimetic control purposes. Such simulations provide excruciating insight into the human metabolic cost of walking and its efficiencies. More information can be gathered by viewing MSNBC video that explains MIT researchers working on robotic joints and modeling and controls process involved in such designs.
3. Future of Humanoid Robots
H+ (Transhumanism) concept
