Quantum computing is an information processing paradigm with the potential to solve certain problems faster than any algorithm running on classical computer architectures. In the next few years, new processors will be developed that support quantum computations exceeding the simulation ability of even the largest classical computer systems. A number of academic and industrial groups are developing prototypes of such devices, also known as NISQ (Noisy Intermediate Scale Quantum) processors. Much as software must be compiled to run on classical computers, quantum algorithms must be compiled to take into account the constraints of particular NISQ devices. Especially in these early prototypes, algorithm performance degrades with runtime due to noise; for this reason, minimizing the runtime of the compiled algorithm (which is represented by a ”quantum circuit”) is critical. We describe a software framework to enable an automated reasoning approach to Quantum Circuit Compilation for NISQ architectures (QCC-NISQ), and our current implementation of it as part of software suite for automated, architecture-aware, compilation for emerging quantum computers. The key components of this suite are a circuit synthesizer, a QCC solver, and a visualizer. These tools provide critical support for the continued development of practical quantum computers and research into quantum algorithms.

Quantum Circuit Compilation: An Emerging Application for Automated Reasoning

Small spacecraft now have precise attitude control systems available commercially, allowing them to slew in 3 degrees of freedom, and capture images within short notice. When combined with appropriate software, this agility can significantly increase response rate, revisit time and coverage. In prior work, we have demonstrated an algorithmic framework that combines orbital mechanics, attitude control and scheduling optimization to plan the time-varying, full-body orientation of agile, small spacecraft in a constellation. The proposed schedule optimization would run at the ground station autonomously, and the resultant schedules uplinked to the spacecraft for execution. The algorithm is generalizable over small steerable spacecraft, control capability, sensor specs, imaging requirements, and regions of interest. In this article, we modify the algorithm to run onboard small spacecraft, such that the constellation can make time-sensitive decisions to slew and capture images autonomously, without ground control. We have developed a communication module based on Delay/Disruption Tolerant Networking (DTN) for onboard data management and routing among the satellites, which will work in conjunction with the other modules to optimize the schedule of agile communication and steering. We then apply this preliminary framework on representative constellations to simulate targeted measurements of episodic precipitation events and subsequent urban floods. The command and control efficiency of our agile algorithm is compared to non-agile (11.3x improvement) and non-DTN (21% improvement) constellations.

Autonomous Scheduling of Agile Spacecraft Constellations with Delay Tolerant Networking for Reactive Imaging

Coalition operations are essential for responding to the increasing number of world-wide incidents that require large-scale humanitarian assistance. Many nations and non-governmental organizations regularly coordinate to address such problems but their cooperation is often impeded by limits on what information they are able to share. In this paper, we consider the     use of an advanced cryptographic technique called secure multi-party computation to enable coalition members to achieve joint objectives while still meeting privacy requirements. Our particular focus is on a multi-nation aid delivery scheduling task that involves coordinating when and where various aid provider nations will deliver relief materials after the occurrence of a natural disaster. Even with the use of secure multi-party computation technology, information about private data can leak.  We describe how the emerging field of  quantitative information flow can be used to help data owners understand the extent to which private data might become vulnerable as the result of possible or actual scheduling operations, and to enable automated adjustments of the scheduling process to ensure privacy requirements

Privacy-aware Adaptive Scheduling for Coalition Operations

When solving real-life problems we often encounter issues that are not captured by academic benchmark domains. In this paper we consider an application problem, representative of a class of real-world problems that have interesting properties: long solution plans with many temporal/numeric constraints.  We identify a number of limitations of a popular family of planners in solving these problems. This family of planners perform Forward Search and call an LP solver multiple times at every state to check for consistency, and to set bounds on the numeric variables in order to determine action applicability. These checks during search allow the pruning of branches; however, they do carry computational cost.  In this paper we investigate and analyse this trade-off, with particular reference to our class of application problems, and show that adapting the planners to call the LP solver less often, and using a cheaper consistency check at each state, can improve performance.

Evaluating the Cost of Employing LPs and STPs in Planning: Lessons Learned From Large Real-Life Domains

Learning-based methods are increasingly popular for search algorithms in single-criterion optimization problems. On the other hand, there are fewer approaches when dealing with multiple-criteria optimization, while numerous applications exist. Constrained path-planning for Autonomous Ground Vehicles (AGV) is one such application, where an AGV is typically deployed in disaster relief or search and rescue applications in off-road situations. The agent can be faced with the following dilemma : optimize a source-destination path according to a known criterion and an uncertain criterion under operational constraints. The known criterion is associated to the cost of the path, which represents the distance. The uncertain criterion represents the feasibility of driving through the path without requiring human intervention. It depends on parameters such as the physics of the vehicle, state of the explored terrains or weather conditions. In this work, we run simulations offline in which we train a neural network model to predict the uncertain criterion. We integrate this model inside a path-planner which can solve problems online. Finally, we carry out experiments on realistic AGV scenarios which suggest the proposed framework requires human intervention less frequently, trading for a limited increase in the path distance.

Learning-based Preference Prediction for Constrained Multi-Criteria Path-Planning

Cloud Migration transforms customer’s data, application and services from original IT platform to one or more cloud en- vironment, with the goal of improving the performance of the IT system while reducing the IT management cost. The enterprise level Cloud Migration projects are generally com- plex, involves dynamically planning and replanning various types of transformations for up to 10k endpoints. Currently the planning and replanning in Cloud Migration are generally done manually or semi-manually with heavy dependency on the migration expert’s domain knowledge, which takes days to even weeks for each round of planning or replanning. As a result, automated planning engine that is capable of gener- ating high quality migration plan in a short time is particu- larly desirable for the migration industry. In this short paper, we briefly introduce the advantages of using AI planning in Cloud Migration, a preliminary prototype, as well as the challenges the requires attention from the planning and scheduling society.

Advantages and Challenges of Using AI Planning in Cloud Migration

We describe the use of an automated scheduling system for observation policy design and to schedule operations of the NASA (National Aeronautics and Space Administration) ECOSystem Spaceborne Thermal Radiometer Experiment on Space Station (ECOSTRESS). We describe the adaptation of the Compressed Large-scale Activity Scheduler and Planner (CLASP) scheduling system to the ECOSTRESS scheduling problem, highlighting multiple use cases for automated scheduling and several challenges for the scheduling technology: handling long-term campaigns with changing information, Mass Storage Unit Ring Buffer operations challenges, and orbit uncertainty. The described scheduling system has been used for operations of the ECOSTRESS instrument since its nominal operations start July 2018 and is expected to operate until mission end in Summer 2019.

Automated Science Scheduling for the ECOSTRESS Mission

Rising demand for home health care services combined with a shortage of professionals leads to increasing workload of existing employees. However, there is a performance limit so that it will no longer be possible to offer these services maintaining quality and economic viability without changing operational management.
To cope with this situation, we propose a concept of task bundle splitting options among several cooperative agents with different qualifications. Further, we integrate this concept in a planning and scheduling algorithm for multiple concurrent agent actions which can  increase efficiency and can improve use of limited resources in operational processes.
For this purpose, possible concurrent actions of agents with different qualifications were evaluated combined with the scheduling process and compared to alternatives. As a first step, this contribution presents the concept as well as an algorithm generating an optimal solution and gives an insight into future work.

Planning and Scheduling for Cooperative Concurrent Agents with Different Qualifications in the Domain of Home Health Care Management

We describe three approaches to enabling an extremely computationally limited embedded scheduler to consider a small number of alternative activities based on resource availability. We consider the case where the scheduler is so computationally limited that it cannot backtrack search. The first two approaches precompile resource checks (called guards) that only enable selection of a preferred alternative activity if sufficient resources are estimated to be available to schedule the remaining activities. The final approach mimics backtracking by invoking the scheduler multiple times with the alternative activities. We present an evaluation of these techniques on mission scenarios (called sol types) from NASA's next planetary rover where these techniques are being evaluated for inclusion in an onboard scheduler.

Enabling Limited Resource-Bounded Disjunction in Scheduling

Generating and scheduling activities is particularly challenging
when considering both consumptive resources and
complex resource interactions such as time-dependent resource
usage.We present three methods of determining valid
temporal placement intervals for an activity in a temporally
grounded plan in the presence of such constraints. We introduce
the Max Duration and Probe algorithms which are
sound, but incomplete, and the Linear algorithm which is
sound and complete for linear rate resource consumption.
We apply these techniques to the problem of scheduling
awakes for a planetary rover where the awake durations
are affected by existing activities. We demonstrate how the
Probe algorithm performs competitively with the Linear algorithm
given an advantageous problem space and well-defined
heuristics. We show that the Probe and Linear algorithms
outperform the Max Duration algorithm empirically.
We then empirically present the runtime differences between
the three algorithms. The Probe algorithm is currently base-lined
for use in the onboard scheduler for NASA’s next planetary
rover, the Mars 2020 rover.

Scheduling with Complex Consumptive Resources for a Planetary Rover

Dynamic ridesharing services (DRS) play a major role in improving the efficiency of urban transportation. User satisfaction in dynamic ridesharing is determined by multiple factors such as travel time, cost, and social compatibility with co-passengers. Existing DRS optimize profit by maximizing the operational value for service providers or minimize travel time for users but they neglect the social experience of riders, which significantly influences the total value of the service to users. We propose DROPS, a dynamic ridesharing framework that factors the riders' social preferences in the matching process so as to improve the quality of the trips formed. Scheduling trips for users is a multi-objective optimization that aims to maximize the operational value for the service provider, while simultaneously maximizing the value of the trip for the users. The user value is estimated based on compatibility between co-passengers and the ride time. We then present a real-time matching algorithm for trip formation. Finally, we evaluate our approach empirically using real-world taxi trips data, and a population model including social preferences based on user surveys. The results demonstrate improvement in riders' social compatibility, without significantly affecting the vehicle miles for the service provider and travel time for users.

ICAPS 2019 Workshop SPARK Submissions