-- Fully autonomous navigation using nano drones has numerous application in the real world, ranging from search and rescue to source seeking. Nano drones are well-suited for source seeking because of their agility, low price, and ubiquitous character . Unfortunately, their constrained form factor limits flight time, sensor payload, and compute capability. These challenges are a crucial limitation for the use of source-seeking nano drones in GPSdenied and highly cluttered environments. Hereby, we introduce a fully autonomous deep reinforcement learning-based light-seeking nano drone. We present the method for efficiently training, converting, and utilizing deep reinforcement learning policies. Our training methodology and novel quantization scheme allow fitting the trained policy in 3 kB of memory. The quantization scheme uses representative input data and input scaling to arrive at a full 8-bit model. Finally, we evaluate the approach in simulation and flight tests using a Bitcraze CrazyFlie, achieving 80% success rate on average in a highly cluttered and randomized test environment. Even more, the drone finds the light source in 29% fewer steps compared to a baseline simulation (obstacle avoidance without source information). T o our knowledge, this is the first deep reinforcement learning method that enables source seeking within a highly constrained nano drone demonstrating robust flight behavior . Our general methodology is suitable for any (source seeking) highly constrained platform using deep reinforcement learning. In recent years, nano drones have gained traction in the robotics community. Their agility, maneuverability, and low price make them suitable for a wide range of applications, especially in GPSdenied and cluttered environments.

This paper addresses two challenges facing sampling-based kinodynamic motion planning: a way to identify good candidate states for local transitions and the subsequent computationally intractable steering between these candidate states. Through the combination of sampling-based planning, a Rapidly Exploring Randomized Tree (RRT) and an efficient kinodynamic motion planner through machine learning, we propose an efficient solution to long-range planning for kinodynamic motion planning. First, we use deep reinforcement learning to learn an obstacle-avoiding policy that maps a robot's sensor observations to actions, which is used as a local planner during planning and as a controller during execution. Second, we train a reachability estimator in a supervised manner, which predicts the RL policy's time to reach a state in the presence of obstacles. Lastly, we introduce RL-RRT that uses the RL policy as a local planner, and the reachability estimator as the distance function to bias tree-growth towards promising regions. We evaluate our method on three kinodynamic systems, including physical robot experiments. Results across all three robots tested indicate that RL-RRT outperforms state of the art kinodynamic planners in efficiency, and also provides a shorter path finish time than a steering function free method. The learned local planner policy and accompanying reachability estimator demonstrate transferability to the previously unseen experimental environments, making RL-RRT fast because the expensive computations are replaced with simple neural network inference. Video: https://youtu.be/dDMVMTOI8KY

Many continuous control tasks have easily formulated objectives, yet using them directly as a reward in reinforcement learning (RL) leads to suboptimal policies. Therefore, many classical control tasks guide RL training using complex rewards, which require tedious hand-tuning. We automate the reward search with AutoRL, an evolutionary layer over standard RL that treats reward tuning as hyperparameter optimization and trains a population of RL agents to find a reward that maximizes the task objective. AutoRL, evaluated on four Mujoco continuous control tasks over two RL algorithms, shows improvements over baselines, with the the biggest uplift for more complex tasks. The video can be found at: \url{https://youtu.be/svdaOFfQyC8}.

Long-range indoor navigation requires guiding robots with noisy sensors and controls through cluttered environments along paths that span a variety of buildings. We achieve this with PRM-RL, a hierarchical robot navigation method in which reinforcement learning agents that map noisy sensors to robot controls learn to solve short-range obstacle avoidance tasks, and then sampling-based planners map where these agents can reliably navigate in simulation; these roadmaps and agents are then deployed on-robot, guiding the robot along the shortest path where the agents are likely to succeed. Here we use Probabilistic Roadmaps (PRMs) as the sampling-based planner and AutoRL as the reinforcement learning method in the indoor navigation context. We evaluate the method in simulation for kinematic differential drive and kinodynamic car-like robots in several environments, and on-robot for differential-drive robots at two physical sites. Our results show PRM-RL with AutoRL is more successful than several baselines, is robust to noise, and can guide robots over hundreds of meters in the face of noise and obstacles in both simulation and on-robot, including over 3.3 kilometers of physical robot navigation.

We study continuous action reinforcement learning problems in which it is crucial that the agent interacts with the environment only through safe policies, i.e.,~policies that do not take the agent to undesirable situations. We formulate these problems as constrained Markov decision processes (CMDPs) and present safe policy optimization algorithms that are based on a Lyapunov approach to solve them. Our algorithms can use any standard policy gradient (PG) method, such as deep deterministic policy gradient (DDPG) or proximal policy optimization (PPO), to train a neural network policy, while guaranteeing near-constraint satisfaction for every policy update by projecting either the policy parameter or the action onto the set of feasible solutions induced by the state-dependent linearized Lyapunov constraints. Compared to the existing constrained PG algorithms, ours are more data efficient as they are able to utilize both on-policy and off-policy data. Moreover, our action-projection algorithm often leads to less conservative policy updates and allows for natural integration into an end-to-end PG training pipeline. We evaluate our algorithms and compare them with the state-of-the-art baselines on several simulated (MuJoCo) tasks, as well as a real-world indoor robot navigation problem, demonstrating their effectiveness in terms of balancing performance and constraint satisfaction. Videos of the experiments can be found in the following link: https://drive.google.com/file/d/1pzuzFqWIE710bE2U6DmS59AfRzqK2Kek/view?usp=sharing.

Learning in environments with large state and action spaces, and sparse rewards, can hinder a Reinforcement Learning (RL) agent's learning through trial-and-error. For instance, following natural language instructions on the Web (such as booking a flight ticket) leads to RL settings where input vocabulary and number of actionable elements on a page can grow very large. Even though recent approaches improve the success rate on relatively simple environments with the help of human demonstrations to guide the exploration, they still fail in environments where the set of possible instructions can reach millions. We approach the aforementioned problems from a different perspective and propose guided RL approaches that can generate unbounded amount of experience for an agent to learn from. Instead of learning from a complicated instruction with a large vocabulary, we decompose it into multiple sub-instructions and schedule a curriculum in which an agent is tasked with a gradually increasing subset of these relatively easier sub-instructions. In addition, when the expert demonstrations are not available, we propose a novel meta-learning framework that generates new instruction following tasks and trains the agent more effectively. We train DQN, deep reinforcement learning agent, with Q-value function approximated with a novel QWeb neural network architecture on these smaller, synthetic instructions. We evaluate the ability of our agent to generalize to new instructions on World of Bits benchmark, on forms with up to 100 elements, supporting 14 million possible instructions. The QWeb agent outperforms the baseline without using any human demonstration achieving 100% success rate on several difficult environments.

A longstanding goal of behavior-based robotics is to solve high-level navigation tasks using end to end navigation behaviors that directly map sensors to actions. Navigation behaviors, such as reaching a goal or following a path without collisions, can be learned from exploration and interaction with the environment, but are constrained by the type and quality of a robot's sensors, dynamics, and actuators. Traditional motion planning handles varied robot geometry and dynamics, but typically assumes high-quality observations. Modern vision-based navigation typically considers imperfect or partial observations, but simplifies the robot action space. With both approaches, the transition from simulation to reality can be difficult. Here, we learn two end to end navigation behaviors that avoid moving obstacles: point to point and path following. These policies receive noisy lidar observations and output robot linear and angular velocities. We train these policies in small, static environments with Shaped-DDPG, an adaptation of the Deep Deterministic Policy Gradient (DDPG) reinforcement learning method which optimizes reward and network architecture. Over 500 meters of on-robot experiments show , these policies generalize to new environments and moving obstacles, are robust to sensor, actuator, and localization noise, and can serve as robust building blocks for larger navigation tasks. The path following and point and point policies are 83% and 56% more successful than the baseline, respectively.

In particular, reinforcement learning (RL) and feedback control can be used to help a robot achieve a goal. Taking advantage of this body of work, this paper formulates general computation as a feedback-control problem, which allows the agent to autonomously overcome some limitations of standard procedural language programming: resilience to errors and early program termination. Our formulation considers computation to be trajectory generation in the program's variable space. The computing then becomes a sequential decision making problem, solved with reinforcement learning (RL), and analyzed with Lyapunov stability theory to assess the agent's resilience and progression to the goal. We do this through a case study on a quintessential computer science problem, array sorting. Evaluations show that our RL sorting agent makes steady progress to an asymptotically stable goal, is resilient to faulty components, and performs less array manipulations than traditional Quicksort and Bubble sort.

Deep neural networks (DNNs) have become the state-of-the-art technique for machine learning tasks in various applications. However, due to their size and the computational complexity, large DNNs are not readily deployable on edge devices in real-time. To manage complexity and accelerate computation, network compression techniques based on pruning and quantization have been proposed and shown to be effective in reducing network size. However, such network compression can result in irregular matrix structures that are mismatched with modern hardware-accelerated platforms, such as graphics processing units (GPUs) designed to perform the DNN matrix multiplications in a structured (block-based) way. We propose MPDCompress, a DNN compression algorithm based on matrix permutation decomposition via random mask generation. In-training application of the masks molds the synaptic weight connection matrix to a sub-graph separation format. Aided by the random permutations, a hardware-desirable block matrix is generated, allowing for a more efficient implementation and compression of the network. To show versatility, we empirically verify MPDCompress on several network models, compression rates, and image datasets. On the LeNet 300-100 model (MNIST dataset), Deep MNIST, and CIFAR10, we achieve 10 X network compression with less than 1% accuracy loss compared to non-compressed accuracy performance. On AlexNet for the full ImageNet ILSVRC-2012 dataset, we achieve 8 X network compression with less than 1% accuracy loss, with top-5 and top-1 accuracies of 79.6% and 56.4%, respectively. Finally, we observe that the algorithm can offer inference speedups across various hardware platforms, with 4 X faster operation achieved on several mobile GPUs.

Swept Volume (SV), the volume displaced by an object when it is moving along a trajectory, is considered a useful metric for motion planning. First, SV has been used to identify collisions along a trajectory, because it directly measures the amount of space required for an object to move. Second, in sampling-based motion planning, SV is an ideal distance metric, because it correlates to the likelihood of success of the expensive local planning step between two sampled configurations. However, in both of these applications, traditional SV algorithms are too computationally expensive for efficient motion planning. In this work, we train Deep Neural Networks (DNNs) to learn the size of SV for specific robot geometries. Results for two robots, a 6 degree of freedom (DOF) rigid body and a 7 DOF fixed-based manipulator, indicate that the network estimations are very close to the true size of SV and is more than 1500 times faster than a state of the art SV estimation algorithm.