The ICRA 2008 was held in Pasadena, California, on May 19-23, 2008. ICRA 2008. The event consisted of three specific challenges with the overall theme of "space robotics". The main goal of the Challenge, however, was to showcase current research being done in all of the disciplines represented at ICRA and, over the coming years, to benchmark the progress that robotics, as a field, is making on real, hard, relevant problems.
This event simulates the exploration of a small area of a planetary surface. There are a number of sub-challenges in the event, and teams should feel free to attempt as many or as few of these as they want. This event is intended to showcase autonomous systems that operate with a minimum of human intervention.
The Environment
The environment for this event is an area approximately 6m by 6m, covered in gravel to a depth of 10cm. The gravel will most likely be somewhere between"pea gravel" and 1/4 inch pieces. Rocks of varying sizes and compositions will be placed on the surface. A simulated lander will be at one end, with one or more ramps extending down to the surface. Robots will ideally start and end all missions on the lander platform. The ramps will be clearly marked with brightly-colored tape, and at least one of them will be free of obstacles. The surface sand will be smoothed between runs, removing all traces of previous robots.
The Event Elements
Onto the surface.
The robot must leave the lander platform, and make it onto the planetary surface. The robot must identify the ramps off of the platform, orient itself appropriately, and navigate down one of the ramps to the surface. The ramps might be obstructed by deflated "air bags", or might not have deployed properly. The system should recognize and deal with these situations appropriately. The operation will be considered successful if the robot manages to get two meters (measured at the closest point) away from the lander. The lander platform will be constructed to autonomously open (perhaps with some human intervention), and to include potential obstacles on the ramps. The event might also be set up to include obstacles at the bottom of the ramps that are either impassible, or must be navigated around.
Data collection.
The robot must autonomously navigate to a remote science station, at the other side of the planetary surface to collect a set of recorded science data, and then return to the lander. The science station and robot are only equipped with short range (simulated) communications, so the robot must be physically close to the science station or lander in order to transfer data. The robot will be given the location of the science station, in some coordinate frame, but will not be given a map of the environment. The goal is to retrieve the data and bring it back to the lander as quickly as possible.
Map the Environment.
The robot should build an accurate metric map of the environment in a fixed time period. The time period starts when the robot leaves the lander ramp, and the map must be built by the end of the time period. Building a map in the sandbox will be particularly difficult, since there will be a lot of wheel slip, and the surface is not planar. We will purposefully put rises and dips into the sand to make mapping challenging.
Extreme navigation.
Part of the sandbox will contain a number of extremely challenging physical obstacles, and a near-vertical cliff-face. Robots must traverse these obstacles to reach the bottom of the cliff, climb the cliff to the top (if possible), and then return to the lander.
Find the robot.
A previous mission has gone wrong, and one of our robots is missing. The goal is to find this robot, following it's tracks in the sand, and to bring it back to the lander. For robots that are incapable of picking up another robot, finding the other robot and touching it will be enough. To add difficulty to this problem, the lost robot can be made small, so that the finding robot cannot just follow the tracks (through narrow spaces), but must plan paths around obstacles, and find the trail again.
Bring back shiny things.
The robot must go out and find interesting objects, and bring them back to the lander. These objects could be unusual rocks, differently colored sand or gravel, alien plants, or man-made artifacts. We could even have the robot look for liquid water or buried treasure.
Back on the lander.
At the end of the mission, the robot must return to the lander platform.
This event simulates an unexpected problem occurring at a planetary habitat, where a robotic solution must be quickly developed and deployed, using only existing resources. The intent of this event is to develop versatile robotic systems and software that can be adapted quickly to address unexpected events. Since humans are present, a natural solution to realistic unexpected events would exploit human creativity and human-robot interaction.
The competition drives not only the development of versatile robotic hardware and on-board software, but also the design and development of programming and assembly tools capable of rapidly implementing a wide variety of capabilities. Since teleoperation is not precluded for this event, the development of effective user interfaces is another expected outcome.
The Environment and Event Parameters
The environment for this event will consist of two areas: the planetary surface and the habitat. The planetary surface will have the same specifications as for the Sandbox event. The habitat will represent the human-occupied structure from which the robots will be "launched" onto the planetary surface. In all scenarios, the human participants cannot exit the habitat. Robots must be placed in an airlock chamber and drive (or be driven) out onto the planetary surface. If a robot needs to return to the habitat, it must do so through the airlock chamber. The airlock will have a pair of sealing doors, making autonomy or wireless teleoperation the only options for robot control. The airlock will be 1.5m long, 1m wide, and 1m tall, with 1m by 1m doors at each end of the long dimension.
Teams will be allowed to use only what they can carry within a container with outside dimensions summing to less than 150cm, and weighing 25kg or less. For example, a container 70cm long, 50cm wide, and 30cm tall has a total dimension of 70+50+30 = 150cm, and would be within the size limits. These limitations are designed to represent the very real space and weight restrictions enforced on space missions, and to make the event more challenging. For convenience, we will also allow access to six standard domestic AC power outlets (United States standard NEMA 5-15, 110v, 15A, 60Hz).
The actual unexpected problem to be solved will be announced on the day of the competition, and can include anything that one can imagine happening on an extra-terrestrial habitation. The problems will be constrained to have likely robotic solution that fit the spirit of the competition. For example, you will not be required to have the robot travel 100km to the site of the problem, or to construct a 10-person emergency habitat from freshly-mined regolith. The scope of the task might vary from a short 1-hour task, to one taking from 4-6 hours. Specific tasks will be announced to all teams simultaneously, and they will work on their solutions independently.
Example Scenarios
To give you an idea of the sorts of tasks that we have in mind, here are a few examples. These tasks are meant to be representative, and to convey the spirit of the competition. We do not guarantee that any of these tasks will actually be assigned during the competition. On the other hand, we do not guarantee that they will not.
Antenna recovery.
An antenna outside the habitat has been knocked over during a Martian storm. The antenna is crucial to the guidance of a resupply transport, which is scheduled to arrive in 4 hours and an EVA is not safe. The team must develop a robot that can reach the antenna, grasp it and reattach it to its receptacle. The antenna is 10 m from the habitat, sitting on top of a 1m by 2m rectangular base that is 1m tall. The base is visible from the habitat. The antenna is a 1cm cylindrical rod 1m long that fits as a peg into a hole 2cm deep in the base. You have a spare antenna and base in the habitat that can be used for testing purposes.
Base station repair.
Sensors have discovered a tear in a thermal covering on the top of storage shed which contains the habitat's store of liquid nitrogen. The team has 4 hours before the Martian morning arrives and starts to dangerously heat the nitrogen. The team must develop a robot that can crawl on top of the structure, use the supplied patching material, and patch the hole by dispensing supplied glue. Unfortunately, the structure was not designed to support heavy weights, so the robot must weigh less than 5kg or risk collapsing the structure, with disasterous consequences.
Buggy Crisis.
HRI has by now become a major research field in robotics. The ICRA'08 HRI Challenge aims at demonstrating a number of state-of-the-art platforms in HRI, as well as provides a realistic platform (the ICRA Conference) for evaluating the effectiveness of the interaction.
Rules and participation
In order to leave the floor open to any team working in HRI, there are no specific requirements neither on the shape and sensori-motor capabilities of the robot, or on the experimental context. We accept any robot: wheel-based platforms, as much as humanoids ones are welcome. We also accept WOZ type of experiments or video-based HRI experiments. And we leave it up to the teams to define the experimental context, i.e. the type of the interaction.
Thus, at this stage, the sole requirements we set for taking part in the HRI Challenge concern the robot's behaviour. The robot should be at least endowed with either of the following two capabilities:
Evaluation:
The effectiveness of a robot engaging in HRI must be evaluated by human users who got the chance to interact with the robot for a sufficiently long period of time. This challenge thus requires that the robots be highly interactive and run constantly throughout ICRA.
The robots' behaviour will be evaluated formally by a team composed of 10 official evaluators (10 experts in robotics, but not necessarily in HRI) and lay people (all the people attending the conference will be given a questionnaire to fill in). The questionnaires will be prepared by experts in HRI evaluation methodologies. The exact scoring system is yet to be defined, but it will at least encompass a score according to the two requirements on the robot's behaviour listed above.
More information can be found in the official website.