Kristofer S.J. Pister (Advisor)

As the post-PC era emerges, several new niches of computer system design are taking shape with characteristics that are quite different from traditional desktop and server regimes. One of the most interesting of these new design regimes is networked sensors. The networked sensor is enabled, in part, by "Moore's Law" pushing computing and storage into a smaller, cheaper, lower-power unit. However, three other trends are equally important: complete systems on a chip, integrated low-power communication, and integrated low-power devices that interact with the physical world. The combination of...

Learning effective policies with model-based reinforcement learning is highly dependent on the accuracy of the dynamics model. Recently, a new parametrization called the trajectory-based model was introduced, which takes in an initial state, a future time index, and control policy parameters, and returns the state at that future time index [3]. This new method has demonstrated improved prediction accuracy in long horizons, increased sample efficiency, and ability to predict the task reward. However, this model has limited transferability to MBRL due to the limited expressivity of its...

The Single Chip Micro Mote (SCμM) is a crystal-free radio chip with an on-board CPU developed at UC Berkeley in the Swarm Lab. This chip was designed to function as the brain of an untethered microrobot with no external components other than a power source required. SCμM also features an optical programmer so that no cables are even needed to program it. With its size and functionality, SCμM has the potential to allow a swarm of microrobots to communicate and perform complex tasks in tandem.

Like every other circuit, SCμM needs power to operate, and one of the most common power...

Some robotic localization methods, such as ultra wideband localization and lighthouse localization, require external localization infrastructure in order to operate. However, there are situations where this localization infrastructure does not exist in the field, such as robotic exploration tasks. Deploying low power wireless sensor networks (WSNs) as localization infrastructure can potentially solve this problem. In this work, we demonstrate the use of an OpenWSN network of miniaturized low power sensor nodes as localization infrastructure. We demonstrate a quadrotor performing laser-...