Jen Jen Chung

Implicit adaptive multi-robot coordination in dynamic environments

Multi-robot teams offer key advantages over single robots in exploration missions by increasing efficiency (explore larger areas), reducing risk (partial mission failure with robot failures), and enabling new data collection modes (multi-modal observations). However, coordinating multiple robots to achieve a system-level task is difficult, particularly if the task may change during the mission. In this work, we demonstrate how multiagent cooperative coevolutionary algorithms can develop successful control policies for dynamic and stochastic multi-robot exploration missions. We find that agents using difference evaluation functions (a technique that quantifies each individual agents contribution to the team) provides superior system performance (up to 15%) compared to global evaluation functions and a hand-coded algorithm.

Evolving memory-augmented neural architecture for deep memory problems

In this paper, we present a new memory-augmented neural network called Gated Recurrent Unit with Memory Block (GRU-MB). Our architecture builds on the gated neural architecture of a Gated Recurrent Unit (GRU) and integrates an external memory block, similar to a Neural Turing Machine (NTM). GRU-MB interacts with the memory block using independent read and write gates that serve to decouple the memory from the central feedforward operation. This allows for regimented memory access and update, administering our network the ability to choose when to read from memory, update it, or simply ignore it. This capacity to act in detachment allows the network to shield the memory from noise and other distractions, while simultaneously using it to effectively retain and propagate information over an extended period of time. We evolve GRU-MB using neuroevolution and perform experiments on two different deep memory tasks. Results demonstrate that GRU-MB performs significantly faster and more accurately than traditional memory-based methods, and is robust to dramatic increases in the depth of these tasks.

Fast Marching Adaptive Sampling

A challenging problem for autonomous exploration is estimating the utility of future samples. In this paper, we consider the problem of placing observations over an initially unknown continuous cost field to find the least-cost path from a fixed start to a fixed goal position. We propose the adaptive sequential sampling algorithm FMEx to successively select observation locations that maximize the probability of improving the best path. FMEx evaluates a set of proposed observation locations using a novel fast marching update method and selects a location based on the probabilistic likelihood of improving the current best path. Simulated results show that FMEx finds lower-cost paths with fewer samples than random, maximum variance and confidence bound sampling. We also show results for sampling bathymetric data to find the best route for a submarine cable. In problems where sampling is expensive, FMEx selects observation locations that minimize the true path cost.

Learning to trick cost-based planners into cooperative behavior

In this paper we consider the problem of routing autonomously guided robots by manipulating the cost space to induce safe trajectories in the work space. Specifically, we examine the domain of UAV traffic management in urban airspaces. Each robot does not explicitly coordinate with other vehicles in the airspace. Instead, the robots execute their own individual internal cost-based planner to travel between locations. Given this structure, our goal is to develop a high-level UAV traffic management (UTM) system that can dynamically adapt the cost space to reduce the number of conflict incidents in the airspace without knowing the internal planners of each robot. We propose a decentralized and distributed system of high-level traffic controllers that each learn appropriate costing strategies via a neuro-evolutionary algorithm. The policies learned by our algorithm demonstrated a 16.4% reduction in the total number of conflict incidents experienced in the airspace while maintaining throughput performance.

Risk-aware graph search with dynamic edge cost discovery

In this paper, we introduce a novel algorithm for incorporating uncertainty into lookahead planning. Our algorithm searches through connected graphs with uncertain edge costs represented by known probability distributions. As a robot moves through the graph, the true edge costs of adjacent edges are revealed to the planner prior to traversal. This locally revealed information allows the planner to improve performance by predicting the benefit of edge costs revealed in the future and updating the plan accordingly in an online manner. Our proposed algorithm, risk-aware graph search (RAGS), selects paths with high probability of yielding low costs based on the probability distributions of individual edge traversal costs. We analyze RAGS for its correctness and computational complexity and provide a bounding strategy to reduce its complexity. We then present results in an example search domain and report improved performance compared with traditional heuristic search techniques. Lastly, we implement the algorithm in both simulated missions and field trials using satellite imagery to demonstrate the benefits of risk-aware planning through uncertain terrain for low-flying unmanned aerial vehicles.

Guest editorial: Special issue on online decision making in multi-robot coordination

Online decisionmaking is an important part of robotics problems in which mobile robots operate in unknown or partially known dynamic environments with the goal of acquiring information about some studied phenomena. This can be found in problems such as autonomous data collection, environmental monitoring, and robotic exploration missions that can be considered as variants of robotic information gathering. The key aspect of these problems is that the overall mission performance can only be evaluated after the mission is completed; however, the choice of which action to take at any time depends on local in-situ conditions that vary according to the information acquired during the mission. This special issue aims at presenting the state-of-the-art in approaches to online decision making for coordinating a team of mobile robots to fulfill a global mission objective through the individual actions of each robot. The particular focus is on missions such as multi-robot exploration, persistent environmental monitoring, and adaptive information gathering. The fundamental challenge of these missions is that little or no information about the environment is known in advance. Therefore, one of the problems that must be addressed is how to trade-off exploration of the unknown parts of the environment to collect new information about the operational environment, and exploitation of the current knowledge acquired so far to improve the mission performance.