Because of communication limits, planetary rovers must operate autonomously during consequent durations. The ability to plan under uncertainty is one of the main components of autonomy. Previous approaches to planning under uncertainty in NASA applications are not able to address the challenges of future missions, because of several apparent limits. On another side, decision theory provides a solid principled framework for reasoning about uncertainty and rewards. Unfortunately, there are several obstacles to a direct application of decision-theoretic techniques to the rover domain. This paper focuses on the issues of structure and concurrency, and continuous state variables. We describes two techniques currently under development that address specifically these issues and allow scaling-up decision theoretic solution techniques to planetary rover planning problems involving a small number of goals.

In over-subscription planning (OSP), the set of goals is not achievable jointly, and the task is to find a plan that attains the best feasible subset of goals. Recent classical OSP algorithms ignore the uncertainty inherent in many natural application domains where OSPs arise. And while modeling stochastic OSP problems as MDPs is easy, the resulting models are too large for standard solution approaches. Fortunately OSP problems have a natural two-tiered hierarchy, and in this paper we adapt and extend tools developed in the hierarchical reinforcement learning community in order to effectively exploit this hierarchy and obtain compact, factored policies. Typically, such policies are sub-optimal, but under certain assumptions that hold in our planetary exploration domain, our factored solution is, in fact, optimal. Our algorithms work by repeatedly solving a number of smaller MDPs, while propagating information between them. We evaluate a number of variants of this approach on a set of stochastic instances of the Rover domain, showing subtantial performance gains.

Abstract — Future planetary rover missions, such as the

We consider the problem of optimal planning in stochastic domains with resource constraints, where the resources are continuous and the choice of action at each step depends on resource availability. We introduce the HAO * algorithm, a generalization of the AO * algorithm that performs search in a hybrid state space that is modeled using both discrete and continuous state variables, where the continuous variables represent monotonic resources. Like other heuristic search algorithms, HAO * leverages knowledge of the start state and an admissible heuristic to focus computational effort on those parts of the state space that could be reached from the start state by following an optimal policy. We show that this approach is especially effective when resource constraints limit how much of the state space is reachable. Experimental results demonstrate its effectiveness in the domain that motivates our research: automated planning for planetary exploration rovers. 1.

Abstract. This paper presents an approach towards probabilistic planning with continuous time. It adopts stochastic concepts for continuous probabilities and integrates them into an HTN-based planning framework. Based on uncertain time durations associated with primitive tasks the time consumption probabilities of non-linear plans can be accumulated and thus an overall probability for a successful execution of complex tasks can be derived that guide the search towards plans with a minimized average value/variance of their overall time consumption. An example from software project planning is used to demonstrate our approach. 1

Abstract. This paper presents an approach towards probabilistic planning with continuous resources. It adopts stochastic concepts for continuous probabilities and integrates them into a STRIPS-based planning framework. The approach enables the construction of plans that are guaranteed to meet certain probability thresholds w.r.t. the consumption of critical resources. Furthermore, the consumption probabilities of multiple resources can be accumulated and thus an overall probability for a successful execution of an aggregate plan can be computed. We extend our approach to HTN-based planning and show how heuristics can be derived that lead to plans with a minimized average value/variance of their overall resource consumption. 1

Abstract — This paper considers the path planning problem for an autonomous vehicle in an urban environment populated with static obstacles and moving vehicles with uncertain intents. We propose a novel threat assessment module, consisting of an intention predictor and a threat assessor, which augments the host vehicle’s path planner with a real-time threat value representing the risks posed by the estimated intentions of other vehicles. This new threat-aware planning approach is applied to the CL-RRT path planning framework, used by the MIT team in the 2007 DARPA Grand Challenge. The strengths of this approach are demonstrated through simulation and experiments performed in the RAVEN testbed facilities. I.

We describe the domain of using an Autonomous Underwater Vehicle to find hydrothermal vents located on the sea floor, and explain some of the difficulties in planning in this domain. We also present a simplified model of the domain, and outline a possible approach for on-line plan generation.

Exploration for robotic mapping is typically handled using greedy entropy reduction. Here we show how to apply information lookahead planning to a challenging instance of this problem in which an Autonomous Underwater Vehicle (AUV) maps hydrothermal vents. Given a simulation of vent behaviour we derive an observation function to turn the planning for mapping problem into a POMDP. We test a variety of information state MDP algorithms against greedy, systematic and reactive search strategies. We show that directly rewarding the AUV for visiting vents induces effective mapping strategies. We evaluate the algorithms in simulation and show that our information lookahead method outperforms the others. 1

In this work, we address the problem of coordinating the distributed execution of plans and schedules by multiple agents subject to a number of different execution uncertainties. The coordination of multi-agent teams in uncertain, dynamic domains is a challenging problem requiring the fusion of techniques from many disciplines. We describe an approach based on the dynamic and selective use of a family of different problem-solving strategies that combine stochastic state estimation with repair-based and heuristic-guided planning and scheduling techniques. This approach is implemented as a cognitive problem-solving architecture that combines (i) a deliberative scheduler, which performs partially-centralized solution repair, (ii) an opportunistic scheduler, which locally optimizes resource utilization for plan enhancement, and (iii) a downgrader, which proactively guides the execution into regions of higher likelihood of success. This paper characterizes the complexity of the problem through examples and experiments, and discusses the advantages and effectiveness of the implemented solution.