For robots to successfully be deployed as human assistants in a variety of applications, it is critical that the robots' controllers and planners are designed with the considerations of both the robots' and humans' abilities and needs. In space applications, where energy is a finite and limiting resource in missions, it may prove necessary to exploit the energy storing-component of series elastic actuators to meet the efficiency needs, while operating in harsh and varied environments. In human-occupied workplaces, robots can only provide the needed support to humans if the robot controller can properly reason about and react to humans' requirements and capabilities. This thesis presents and assesses strategies to address these kinds of scenarios. In Chapter 2, we present a trajectory optimization scheme based on sequential linear programming to leverage the energy-storing capabilities of series elastic actuators for high-performance tasks. We discuss the current limitations in optimization strategies for series elastic actuated robots. One of the difficulties of this planning problem is respecting all relevant, low-level actuator constraints and handling system nonlinearities in a computationally efficient manner. Our simulation and hardware experiments demonstrate the leveraging of compliance for faster motions as compared to those that are achieved by the compliant systems' rigid counterparts. Chapter 3 addresses the need for reactive synthesis to be employed to automatically devise human-aware robot controllers for scenarios in which humans and robots continuously collaborate. Through this approach, it is possible to synthesize high-level control policies that are formally guaranteed to meet human requirements. We present a case study in which a robot seeks to deliver work to a human so that the human is productive, but not stressed by her work backlog. We demonstrate the achievement of a human productivity-informed controller using a mobile manipulator robot that picks up and delivers work based on work backlog. One of the challenges of this problem is devising human productivity models that are practical and accurate. We explore a toy scenario in the hope that this research will introduce methodologies that can be generalized for more practical cases