UC Berkeley

Transforming robots from laborers to collaborators promises to significantly broaden their societal impact, but is presently limited by, among other factors, technical feasibility. Interactive robots seek to achieve safe and productive behavior in non-deterministic settings by realizing reactive behavior, which can enable direct human-robot interaction, promising new modalities of power and information flow between human and robots. This can enable assistive or rehabilitative robots which restore or augment human's physical capabilities, as well as expand robotic roles in manufacturing contexts.

However, several novel considerations must be made in analysis of robots which seek to achieve interactive behavior. For physical interaction, guarantees of safety become both more important and harder to demonstrate. Often, complex hardware is introduced to meet interactive design criteria (e.g. joint torque sensors which introduce joint flexibility). These more complex dynamics introduce additional sources of model uncertainty, and are typically accompanied by hierarchical controllers which further obfuscate both safety and performance. Additionally, interactive robots will couple with unmodeled environments, changing the effective dynamics of the robot and presenting further analytical challenges.

This dissertation examines the safety and performance of uncertain, interactive systems from several perspectives. Limitations to achievable model accuracy and the effects of this model uncertainty on performance and safety are examined analytically and experimentally on series-elastic actuated systems.

First, the objectives of interactive robots and constraints introduced by their hardware are introduced. A model for interactive flexible joint robots is motivated which explicitly considers backdriveability of the motor and load-side dynamics. Conditions for passivity of flexible-joint robots which render a load-side impedance are developed, then extended to hold over an uncertain motor model. This robust passivity condition is shown to induce sensible constraints on inner-loop torque controllers. Rigorous means of relaxing this passivity condition are introduced, and the relaxed condition shown to explain interactions known in literature to be unsafe in practice. A model and corresponding uncertainty bound of an experimental setup is characterized through bilateral system identification (i.e. both motor and environment driven) and the results used to validate the robust passivity condition as a practical design tool. This analytical methodology assists hierarchical controller synthesis by generating practical constraints on controller parameters.

Performance of a linear interactive system is then defined, and shown to be limited by model uncertainty. However, by incorporating direct measurement of interactive variables (force and motion), this performance can be improved and the robust rendering of desired dynamics can be achieved. A novel controller structure, derived from the disturbance observer, is proposed and analyzed. Again, conditions for passivity are developed, then extended to hold over an uncertain model. For a fixed-structure controller, these conditions are then propagated back onto parameter constraints to inform controller design. The performance of this approach is then validated experimentally.

When direct measurement of the interactive force is not feasible, performance improvements can be made by improving model accuracy. Here, a data-driven modeling technique is used to describe non-idealized dynamics. However, interactive systems will couple with unknown environments, making their effective dynamics multimodal and potentially introducing unknown input which confounds identification. A modeling approach suited to these challenges is introduced which allows for the identification of inverse dynamics which are multimodal and subject to intermittent unobserved external disturbances. The passivity of the overall resulting controller policy is shown, and the performance and passivity are validated experimentally.

Uncertainty in the environment motivates the need for interactive control, but realization of this control is in turn limited by uncertainty in the robot model. Explicitly exploring the relationship between model uncertainty, safety, and performance can allow relevant limitations to be improved when possible, and respected when not.