Effectiveness of cooperative and competitive shared control

Advances in technology place ever increasing demands for effective interactions between humans and machines. Human-machine interaction (HMI) that incorporates shared control, in which the human and machines both simultaneously influence the outcome, may lead to a more natural interaction between people and machines. This natural interaction could be particularly beneficial in assistive devices that are used to increase, maintain, or improve capabilities of individuals. An interactive computer simulation of an inverted pendulum which takes input from artificial and human controllers was programmed in Matlab to determine the effectiveness of cooperative shared control. A proportional-derivative (PD) controller was used as the artificial/computer side of the shared control. Input from a human operator was obtained using an Xbox 360 controller, with biofeedback provided by a flat panel display. The artificial controller and human worked together to balance the inverted pendulum vertically and prevent it from falling below the horizontal axis. Random perturbations were provided to destabilize the system. The amount of time in which a participant could maintain stability was used as a performance measurement. In competitive shared control the computer assists the human in completing the primary task. However, in addition, the human works to achieve a secondary task while working symbiotically with the artificial controller. This may result in conditions where the human is competing with the artificial controller to achieve different goals. Note that this type of shared control is different than the "winner takes all" competitive control because influence from each source is always present. The amount of time the pendulum is balanced and how long the pendulum remained within the target area was used as a performance measurement. A total of 20 participants for the cooperative shared control and 12 participants for the competitive shared control were evaluated at 26 different testing conditions in a pseudo-randomized order. Each test condition was repeated three times for each participant and the result for each test condition was averaged. The results from both the cooperative and competitive shared control testing were very promising. The results showed that blended shared control can outperform a human and that higher performance can be achieved by increasing the PD level. Blended shared control can also perform better than an artificial PD controller alone when the difficulty increases beyond the controller's capabilities. This same observation can be made when comparing blended shared control to additive performance. Competitive testing was also able to show that giving the human a secondary task to complete did not interfere with primary task completion. By lightening the load of a primary task, blended shared control could enable someone to perform additional tasks or allow them to perform them better than they could on their own.