The best known defence system with au tonomous func tionalit y currently deployed by the Armed Forces is the Active Protection System (APS) for armoured vehicles, which autonomously destroys incoming anti-tank missiles, rockets or projectiles. To be able to do that, APS combines either radar-based or infrared (IR) sensors which detect incoming projectiles with a fire control system that can track, evaluate and classify threat scenarios.

The entire process, from detection to tracking and engagement, is fully automatised as human intervention would only slow it down or make a timely response impossible altogether. Human operators simply couldn’t act quickly enough to authorise or even supervise the required response. However, APSs are always pre-programmed in such a way that users can anticipate the exact circumstances under which the system will have to engage and respond, and in which cases it shouldn’t. The type of threats that will trigger an APS response are known in advance or at least predictable with a high degree of certainty.

Similar principals also determine the functioning of other autonomous land weapon systems like Counter Rocket, Artillery and Mortars (C-RAM) systems used to protect military bases in war zones. Both APS and C-RAMs can thus be considered as autonomous systems which, once activated, do not require human intervention.


A challenge: autonomy for unmanned ground vehicles

To date, unmanned ground systems are usually used for explosive detection and disposal or reconnaissance of terrains or buildings. In both cases, robots are tele-operated and remotely controlled by human operators (although some robof certainty. ts could perform simple tasks like point-topoint movement without constant human help). “The reason why human intervention remains crucial is that unmanned ground vehicles face tremendous difficulties when operating autonomously in difficult and unpredictable terrain. Having a vehicle moving autonomously on a battlefield where it has to circumvent obstacles, cross moving objects and face enemy fire is much more complex – due to unpredictability – than using an autonomous weapons system such as the afore-mentioned APS”, says Marek Kalbarczyk, EDA’s Project Officer Land Systems Technologies. Therefore, autonomy of unmanned ground vehicles is today still limited to simple functions like ‘follow me’ and waypoint navigation. The ‘follow me’ function can be used either by unmanned vehicles to follow another unmanned/manned vehicle or a soldier, while waypoint navigation allows a vehicle to use the co-ordinates (as defined by an operator or learnt by the system) to reach its desired destination. In both cases an unmanned vehicle uses GPS, radar, visual or electromagnetic signatures or radio links to follow the lead vehicle or defined/learnt path.


Soldier protection

From an operational point of view, the objectives for using such autonomous functions are usually to:

  • decrease exposure for soldiers in dangerous zones by replacing drivers with unmanned vehicles or driverless kits with autonomous following function in convoys, or
  • provide support to troops in remote areas.

Both functions commonly rely on a so-called ‘avoid obstacle’ feature to prevent collisions with obstacles. Due to the complex topography and shape of certain land areas (hills, valleys, rivers, trees, etc.), the waypoint navigation system used in land platforms has to include LiDAR (Light Detection And Ranging) capabilities, or be able to use pre-loaded maps. However, since LiDAR relies on active sensors, and therefore is easy to detect, the research focus is now shifting towards passive vision-based systems. Pre-loaded maps are sufficient though when unmanned vehicles operate in well-known environments for which detailed maps are already available (for instance when used to monitor and protect borders or critical infrastructures).

The increased use of unmanned and autonomous land systems will require changes in the military educational system too, to properly train system operators

However, LiDAR is imperative for waypoint navigation every time unmanned vehicles have to enter complex and unpredictable environments. The problem is that LiDAR has its limits too, i.e. that its reliability can only be guaranteed for unmanned vehicles operating in semi-complex terrain.

Hence the need for further research and development in this domain. To that end, several technology demonstrators have been developed – for example ADM-H or EuroSWARM – with a view to exploring, testing and demonstrating more advanced functions, including autonomous navigation or cooperation of unmanned systems. These demonstrators, however, are still at an early research phase. 


Many challenges ahead

Limited LiDAR is not the only challenge that unmanned ground vehicles are facing. According to the Unmanned Ground Systems Landscaping and Integration Study (UGS LIS), funded by the European Defence Agency (EDA), as well as another EDA-financed study on the ‘Identification of all major technical and safet y requirements for militar y unmanned vehicle to operate in combined manned-unmanned mission’ (SafeMUVe), the challenges and opportunities can be divided into five different categories:

1. Operational: There are plenty of potential missions that can be envisaged for unmanne d ground vehicles wi t h autonomous functions (communication node, area surveillance, zone and route reconnaissance, casualty extraction, CBRN reconnaissance, follower mule, convoying for the distribution of supplies, route clearance, etc.), but operational concepts to back and underpin these are still lacking. It is therefore difficult for developers of unmanned ground vehicles with autonomous functions to develop systems which for sure will meet military requirements. Creating a forum or working group of defence users of unmanned ground systems with autonomous functions could solve this problem.

2. Technical: The potential benefits of unmanned ground vehicles with autonomous functions are considerable, but so are the technical hurdles still to overcome. Depending on the envisaged mission, unmanned ground vehicles can be equipped with different payload suites (sensors for ISR or CBRN monitoring and detection, manipulators for explosives handing or weapon systems, navigation and guidance systems...), intelligence kits, operator control suites and control hardware. This means that several enabling technologies, such as decision making/ cognitive computing, human machine interaction, computer vision, state of battery technologies or collaborative intelligence, are absolutely critical. In par ticular, unstructured and contested environments pose huge challenges to both navigation and guidance sensors. Here, the way forward has to include the development of new sensors (quantum positioning, ultra-cold atom interferometers, smart G&C actuators…) and techniques such as decentralised and cooperative SLAM (Simultaneous Localization and Mapping) and 3D mapping, relative navigation, advanced hybridisation and data-fusion of available sensors as well as vision/IR aided mobility. The problem is not so much of a technological nature – because most of these technologies are already used in civilian applications – but rather of a regulatory order. Indeed, such technologies cannot immediately be used for military purposes as they need to be adapted to specific military requirements. Against this backdrop, the EDA’s Overarching Strategic Research Agenda (OSRA) is a tool that can deliver this missing piece. Under OSRA, several so-called Technology Building Blocks (TBBs) are being developed, addressing technology gaps related to unmanned ground vehicles, for instance: – Manned/unmanned teaming, adaptive cooperation between man and unmanned system with different levels of autonomy; Health and usage monitoring; Novel User Interfaces for Soldier (assets integration/ control); Nav igation in GNS S denied environment; Autonomous and automated GNC and Decision Making techniques for manned and unmanned systems; Multirobot Control and Cooperation; Precision guidance and control of weapons; Active imaging systems; Artificial Intelligence and Big Data for Decision Making Support. Each TBB is owned by a dedicated panel (called CapTech) composed of governmental, research and industrial experts. Each CapTech will develop a roadmap for each TBB.

3. Normative/Legal: An important obstacle for the introduction of autonomous systems in defence is the lack of suitable verification and evaluation procedures or certification processes which can be used to prove that even the most basic unmanned ground vehicle with autonomous functions is able to operate correctly and safely, even in hostile and complex environments. In the civil domain, self-driving cars are facing the same problems. According to EDA’s SafeMUVe study, the main gaps identified with respect to specific standards/best practices are concentrated in the modules related to the higher layer of autonomy, namely ‘Automation’ and ‘Data fusion’ aspects. Modules such as ‘Environment perception’, ‘Localization & mapping’, ‘Supervision’ (Decision-making), ‘Motion planning’, etc. are still in mid technology readiness levels (TRLs), and although several solutions and algorithms exist to perform the different tasks, no standards are yet available. In this sense, there is also a gap in terms of verification and certification of these modules, partly addressed by the EU funded European Initiative to Enable Validation for Highly Automated Safe and Secure Systems (ENABLE-S3) project. The recent establishment, by the EDA Steering Board, of a Land Test Centres Network of Excellence (LTE) is a first step in the right direction. The LTE allows national test centres to undertake joint initiatives in view of preparing the testing of future technologies, such as automotive systems and robotics.

4. Personnel: The increased use of unmanned and autonomous land systems will require changes in the military educational system too, to properly train system operators. Especially, military staff need to understand the technical principals of a system’s autonomy to ensure they can properly operate and control it when necessary. Building trust between a user and an autonomous system is a prerequisite for the wider use of unmanned ground systems with higher autonomous functions.

5. Financial: Whereas commercial global players like Uber, Google, Tesla or Toyota are investing billions of Euros in developing self-driving cars, military spending on unmanned ground systems is much more modest, and fragmented too, as Member States have their own national development plans. The EU Preparatory Action on Defence Research (PADR) and the future European Defence Fund should help consolidate funding and support a European collaborative research approach to develop unmanned ground systems with more advanced autonomous functions. EDA work EDA has already been active in the unmanned systems land domain for some time. Specific technology aspects such as mapping, path planning, vehicle following, or obstacle avoidance were developed in collaborative research projects like SAM-UGV or HyMUP, both of them jointly funded by France and Germany. The SAM-UGV project aimed to develop an autonomous technology demonstrator based on a mobile land system platform and characterised by a modular architecture both in hardware and software. In particular, the technology demonstrator proved the concept of scalable autonomy (switching between tele-operation, semi-autonomous and autonomous behaviour). The SAM-UGV project was further developed under the HyMUP project which proved the feasibility of mounted combat missions of unmanned systems, in coordination with regular manned vehicles. Additionally, the protection of autonomous systems against enemy interference, safety requirements for combined mannedunmanned mission and the standardisation of UGVs are currently being addressed in EDA’s PASEI project, as well as in the SafeMUVe and SUGV studies respectively.