Autonomous control systems onboard planetary rovers and spacecraft benefit from having cognitive capabilities like learning so that they can adapt to unexpected situations in-situ. Q-learning is a form of reinforcement learning and it has been efficient in solving certain class of learning problems. However, embedded systems onboard planetary rovers and spacecraft rarely implement learning algorithms due to the constraints faced in the field, like processing power, chip size, convergence rate and costs due to the need for radiation hardening. These challenges present a compelling need for a portable, low-power, area efficient hardware accelerator to make learning algorithms practical onboard space hardware. This paper presents a FPGA implementation of Q-learning with Artificial Neural Networks (ANN). This method matches the massive parallelism inherent in neural network software with the fine-grain parallelism of an FPGA hardware thereby dramatically reducing processing time. Mars Science Laboratory currently uses Xilinx-Space-grade Virtex FPGA devices for image processing, pyrotechnic operation control and obstacle avoidance. We simulate and program our architecture on a Xilinx Virtex 7 FPGA. The architectural implementation for a single neuron Q-learning and a more complex Multilayer Perception (MLP) Q-learning accelerator has been demonstrated. The results show up to a 43-fold speed up by Virtex 7 FPGAs compared to a conventional Intel i5 2.3 GHz CPU. Finally, we simulate the proposed architecture using the Symphony simulator and compiler from Xilinx, and evaluate the performance and power consumption.

CubeSats and small satellites are emerging as low-cost tools for performing science and exploration in deep space. These new classes of satellite exploit the latest advancement in miniaturization of electronics, power systems, and communication technologies to promise reduced launch cost and development cadence. JPL's MarCO CubeSats, part of the Mars Insight mission, will head on an Earth escape trajectory to Mars in 2018 and serve as communication relays for the Mars Insight Lander during Entry, Descent and Landing. Incremental advancements to the MarCO CubeSats, particularly in propulsion and GNC, could enable these spacecraft to get to another planet or to Near Earth Objects. This can have substantial science return with the right science instrument. We have developed an interplanetary CubeSat concept that includes onboard green monopropellant propulsion system and that can get into a capture orbit around a neighboring planet or chase a small-body. One such candidate is the Martian moon Phobos. Because of the limits of current CubeSat hardware and lack of an accurate ephemeris of Phobos, there will be a 2 to 5 km uncertainty in distance between the spacecraft and Phobos. This presents a major GNC challenge when the CubeSat first attempts to get into visual range of the moon. One solution to this challenge is to develop optical navigation technology that enables the CubeSat to take epicyclic orbits around the most probable location of the target, autonomously search and home-in on the target body. In worst-case scenarios, the technology would narrow down the uncertainty of the small-body location and then use optical flow, a computer vision algorithm to track movement of objects in the field of view. A dimly lit small-body can be detected by the occlusion of one or more surrounding stars. Our studies present preliminary simulations that support the concept.