Goal-oriented autonomous controller (GOAC)

Abstract

GOAC is a general-purpose autonomous controller that acts as a generic platform that is applicable to a wide range of robotics applications. The space platform, developed for ESA robotic missions, provides artificial intelligence (AI) capabilities to the robot. The system is based on the combined “execution-with-planning” concept. It is constructed over the GenoM robotics framework, enhanced with correct-by-construction techniques.

The technology is suitable for all robotic systems (including terrestrial) in which a greater level of autonomy is required/needed.

Description

The Goal-Oriented Autonomous Controller (GOAC) is based on an agent-oriented architecture. GOAC is a hybrid architecture consisting of a set of reactors/timeline planners over a reactive/interruptible functional layer.

The architecture is illustrated in the figure. Each reactor is a “sense-plan-act” control loop working at different levels and making decisions over different functional aspects of the system. Reactors are defined based on whether they need to “deliberate” on actions at the highest level, or respond to the inputs from the lower levels (closer to the hardware). There is a well-defined messaging protocol for exchanging facts and goals between the reactors:

  • observations of the current state (either from the environment or from within the platform)
  • goals to be accomplished

The intelligence of the controller is in the planner, i.e., the problem solver. However in GOAC, instead of a single planner there can be several planners where each is embedded into a different reactor. The controller can be seen as a network of reactors, where the output of each reactor (a plan) is the input (a set of goals) for another reactor.

Therefore, GOAC follows a divide-and-conquer approach to complexity, by splitting the problem into sub-problems, thus making it more scalable and efficient. The number of reactors needed “to deliberate” can be adapted to each given system and each given mission.

Each reactor deliberates over a different part of the system; for instance, a science reactor and a navigation reactor take into account different aspects of the mission. A mission-level reactor could consider the whole mission life (e.g., a day-long survey). A navigation reactor however would look into the future for a single navigation event, such as a movement towards a given target point.

The problem planning can be computationally intensive, but by splitting it into several sub-problems, a scale of efficiency can be achieved. In one example, if a lower-level reactor can cope with a re-planning requirement then the higher-level reactors will not be informed. In this way, a smaller section of the overall planning system can be used to efficiently complete a re-planning need.

Making decisions in GOAC relies on timeline-based planners. According to this technology, plans are flexible so that the start, duration and end of the planned tasks are not fixed. In this way, plans are more flexible and robust for uncertain environmental conditions, when compared to a predefined, rigid sequence of activities.

Innovations and advantages of the offer

GOAC responds to the needs of robot autonomous missions in which higher degrees of on-board autonomy are required.

By combining different areas in Artificial Intelligence, GOAC provides an operational concept in which mission operators can focus on what they want the robotic platform to do instead of laboriously working on how to satisfy science and engineering goals.

In addition to potentially reducing the costs of the mission, a high-level of autonomy also improves the performance of the robotic platform in two ways:

Scientific performance: high-level guidelines for nominal, but also “opportunistic” scientific exploration, allows greater science efficiency and higher potential for greater mission returns, avoiding an over reliance on pre-calculated and potentially restrictive plans.

Robustness of the robotic system: when a robotic system reaches an “out of mission” situation it can be safer and more efficient to autonomously react to the environmental conditions and autonomously re-plan.  This may be better than waiting for ground instructions or relying on fixed pre-programmed alternative plans.  Greater autonomy in decision making potentially increases the probability of successfully meeting the mission’s objectives.

Further Information

The GOAC controller has been developed under ESA (European Space Agency) activities.

Application

Autonomous robots for terrestrial conditions, including:

  • Underwater exploration
  • Oil and gas platforms
  • High radiation environments
  • Military/security

Comments on the technology by the broker

The GOAC control system is currently used in the FOXIRIS robot (TOTAL Argos Challenge).  See http://esa-tec.eu/success-stories/from-space/robots-under-test-for-oil-and-gas-rig-duty

Description of Space Heritage

Technology developed under ESA contract: (ESA 22361/09/NL/AR)

Category
Automation & Robotics
Reference No.
TDO0206
Could this technology benefit your business? Please contact Richard Seddon Tecnalia (Spain)
Discussion
comments powered by Disqus