WP3 High Accessibility Unmanned Vehicles
This work package will start by examining the options for unmanned vehicles capable of carrying out reconnaissance in an area of the mine where trapped personnel could be located – perhaps through rock falls or small exploratory tunnels – before committing rescuers to the region. Flying, climbing and snake-like crawling robots will be studied, plus a swarming configuration to allow several vehicles to cover large areas of a mine in a reasonable time. This will be followed by developing such a vehicle, utilising communications technologies developed in WP1 and sensors developed in WP2.
The work is divided into the following tasks:
T3.1 – Analysis and Selection of Locomotion Method
Several unmanned vehicle technologies will be investigated in this task, with a view to making an informed decision on which to develop further These are differentiated by different methods of locomotion and can be described as flying (UAVs), climbing and snake-like (crawling), the pros and cons of which are as follows.
Flying vehicles have several advantages such as fast navigation and good incident area observation. Moreover, their potential low cost makes them very suitable for the swarm configurations and substitution in the case of accidents. Nevertheless, their ability to autonomously fly in a complex and confined rescue area is their main drawback.
Climbing vehicles have a particular advantage in their ability to climb in the rescue environment using special types of “legs” that allow them to walk on the ceiling or walls of the tunnels and also over rubble or parts of the mine’s supporting structure. There are several locomotion configurations such as four or six active legs, the locomotion control strategy being bio-inspired (e.g. ants). The attachment mechanism is the crucial issue: suction pumps, magnets and ever glue attachment (dragon-fly) options are possible depending on the environment. The evaluation of this class of vehicle will concentrate particularly on the pros and cons of these options.
Snake-like (crawling) vehicles are especially suitable for extremely complicated environments. Such vehicles, comprising a high number of segments, have the advantage that they can crawl over the rocks and even in flooded conditions. Moreover, due to their modular vehicle concept, segments can easily be added or removed, thereby reconfiguring their kinematics. The main drawback of this type of vehicle is the limited area knowledge, comprising a local view only, the understanding of the global rescue scene being complex and sometimes not possible.
In addition to appraising the individual types of vehicle, a combination of several types of vehicle will be also analysed. Combining vehicles with a high level of global situations understanding (flying) with vehicles with very local data (snake-like) could, potentially, prove especially valuable. However, close attention will be paid to overall control and maintenance, two factors that could, potentially, be jeopardised by poor standardisation when using several technologies of vehicles.
For evaluation and final selection of the most adequate vehicle configuration, the bottom up methodology will be used. First simulation of the mine environment and of the different vehicles types will be carried out using the Gazebo robot simulator with several tools and a human-machine interface that will be developed for the project. Second, an experimental selection will be carried out to include a consideration of robot paths, number of attempts, presence/absence of humans, etc. Third, metrics will be collected for quantification of motion (e.g. time spent, energy consumption, level of manoeuvrability, etc.). Finally, a selection will be made, taking into account not only the quantified data but also several technical aspects including communications, operating system, ease of sensor integration, level of autonomy/tele-operation, etc., and cost, both in terms of the price of the equipment and the operational cost.
T3.2– Integrated Positioning System
An Integrated Positioning System (IPS) for drones has been developed by DLR in collaboration with DMT; it combines a stereoscopic camera system and inertial sensors. A trajectory is calculated from optical and inertial data with an average 3D accuracy in the range of 0.1% of the travelled path. The calculated coordinates and the images can be fitted directly into the original mine plan by special software tools, which have to be developed. The aim of this task is to redesign / adapt the IPS to this very specific application of use on the small, unmanned exploratory vehicles that will be developed in this work package. Development will be carried out to integrate it into the small unmanned vehicles that will be developed in T3.4.
First, the IPS will be downsized to fit the small remote-controlled reconnaissance vehicles. If airborne systems / drones as rescue vehicles are taken into consideration, weight optimization is also an important aspect. One special constraint is, of course, the lack of natural illumination, so this downsized IPS will use very sensitive cameras and very wide angle lights. A good camera resolution will increase the positioning accuracy because smaller structures can be recognized.
In order to get a real time position on a mine plan, special interfacing technology and software will be added. A low-speed channel for the proposed LF/VLF communication will allow the permanent transmission of the current rescue vehicle position and the transmission of rough images at specific time intervals, related to a fixed position. The complete trajectory, with the corresponding image data, will be accessible via a high-speed channel. This high-speed channel may be used only close to the base station due to the in-mine RF propagation conditions and the selected transmission system. In some cases, the first opportunity to download the image data will be after the recovery of the rescue vehicle.
In addition to the above redesign, the IPS will be ruggedized to operate in mining conditions. A series of tests will be carried out to investigate the influence of dust, smoke, fog and strong vibrations. With respect to future projects, e.g. ATEX certification, the power requirement and the voltage level will be kept as low as possible.
As mentioned under T2.3, other environmental sensors will need to be implemented. Rescue teams need to know the atmospheric situation such as temperature, methane concentration and carbon monoxide or carbon dioxide concentrations, as well as seismic activity and acoustic phenomena, for example human voices. Consideration will be given to including these detectors in the IPS and incorporating their data into the IPS data stream. This will result in the creation of a “rescue sensor” which can be attached to various remote controlled or autonomous vehicles. This will make use of the sensor technologies researched in WP2.
T3.3 – Autonomous Navigation of Unmanned Vehicle
Autonomous navigation of unmanned vehicles in a constellation (swarm) is a very challenging task in complex environments. The environmental conditions are very complicated due to the lack of light or navigational beacons, absence of a GPS signal, and the lack of conventional communications. The area is also unstructured, in the sense of not knowing the exact geometric data of the environment for vehicle navigation. The IPS described in T3.2 will be employed both for positioning (exact absolute and relative) of vehicles in the environment, and determining the location of any trapped of injured miners.
The most common technique for area exploration and navigation is referred to as SLAM (Simultaneous Localisation and Mapping) and will be employed in this task. It uses statistical prediction techniques (mainly Kalman filtering) for estimating the vehicle position in the environment and, at the same time, the 3D reconstruction of the mine area. A means will be developed to allow navigation planning and re-planning to be achieved through learning procedures.
Referring to the vehicles’ swarming capability (possibly employing from 3 to 5 units) the state-of-the-art formation techniques are as follows: (1) follow the leader (the leader is always the same and the followers communicate with the leader only), (2) auto-configuration (each vehicle communicates with the closer ones in order to dynamically form different configurations) and (3) randomization (each vehicle randomly navigates and takes data which is mixed using a Gaussian mixture model). All of these options will be studied with a view to selecting the option most suitable for the mine environment.
For the evaluation of the developed algorithm, several techniques and tools will be used. First, the simulator implemented in T3.1 (with a specific module for the IPS sensors from T2.3) will be used in order to evaluate the implementation of SLAM techniques. A MATLAB plug-in will also be used for algorithm adjustment. In the second stage, the same configuration (Gazebo-MATLAB) will be used for swarm technique implementation and evaluation. Different scenarios will be implemented and tested. These include cluttered/free space, presence/absence of humans, deep/shallow mines, etc.
T3.4 – Unmanned Vehicle Development and Prototyping
Building on the results of the previous tasks in this work package, the unmanned vehicle(s) will be designed and prototyped.
Initially, a decision will be made regarding whether the vehicle will be built from scratch, which would be the more expensive solution, or whether commercially available hardware can be used. Also, a mix of both solutions will be considered, i.e. basing the design on a commercial vehicle and carrying out the necessary adaptation, e.g. for ATEX compliance. During this task the following vehicle features will be optimised: minimisation of the overall weight, maximisation of the payload (for sensors and communication equipment), minimisation of the power consumption, high level of manoeuvrability, and immunity to harsh conditions. Throughout this process, initial results of the modelling work carried out in Task 5.3 into environmental conditions following a major incident will be analysed.
An important decision to be made in this task concerns ATEX certification. It is envisaged that the basic parts of the UAVs (motors and their drives, autopilot, communications and battery) will be ATEX compliant. However, employing a conventional ATEX design techniques for the vehicles in their entirety would jeopardise the key aims of extremely small size and weight for rapid and early deployment. Accordingly, less critical systems will be designed to automatically power down, under the control of equipment developed in T2.3, if an explosive atmosphere is detected. The decision on the protection modes to be used for the non-essential subsystems will be made during the project on a case-by-case basis, basing the assessment on the cost of implementation (both financially and in terms of added weight) and performance.
A prototype vehicle will be tested in laboratory conditions, firstly in an open-space area and then in an in-door environment. The navigation can be checked through several path types such as straight line motion, exact positioning, etc. In the next stage, a second vehicle will be used in order to check the vehicles’ swarming capabilities (communication, navigation and formation). For these preliminary tests careful planning is needed. This will lead to underground field testing in WP6.