Periodic Reporting for period 3 - ILIAD (Intra-Logistics with Integrated Automatic Deployment: safe and scalable fleets in shared spaces)
Período documentado: 2019-07-01 hasta 2021-06-30
ILIAD is driven by the application needs for fleets of robots that operate in warehouse logistics applications with a high demand on flexibility, in environments shared with humans. In particular, the project aims to enable automatic deployment of a fleet of autonomous forklift trucks (AGVs), which will continuously optimise its performance over time by learning from collected data. Companies of all scales, from small enterprises to international corporations, are in need of robotic solutions that can integrate with current warehouses, thus facilitating a transition to automation. Small-scale end users with existing warehouses who do not have the means to build new, fully automated, warehouses will require smart automation that can effortlessly integrate with current facilities and efficiently and safely interact with staff. To this end, a system that is easily scalable is required: the overhead deployment cost of the first truck should be minimal, and additional trucks should seamlessly integrate with the existing fleet. More large-scale companies may invest in automated goods-to-person solutions that automatically deliver boxes to human pickers for preparation of customer orders. However, these systems lack the flexibility of traditional warehouses and cannot handle, e. g., oversized objects and dangerous goods. Therefore, solutions like those targeted by the ILIAD project will be required to efficiently handle the remaining fraction of goods, and thus continue the transition to automation for end users of all scales.
The overarching goal of ILIAD is to address limitations in the state of the art which impede the efficient use of robot fleets in warehouse logistics. We address these limitations by the following means: a systematic study of human safety in shared environments, the development of a generic, safe and efficient solution for a mixed fleet of robots handling logistics tasks in human–robot shared environments, supporting life-long operation (meaning that the system can run independently, even when the environment changes), efficient methods for fleet coordination (of mixed fleets of autonomous and human-driven vehicles), and automated picking and handling of a wide range of goods without replacing the gripper.
In order to drive the proposed research and innovations, and to maximise the impact of these actions within the industry, ILIAD has adopted a particularly demanding use case for logistics in the distribution of food products, involving AGVs operating together with human workers. For ILIAD, the food industry provides an especially relevant use case because of its particularly challenging requirements: sensitive products with short shelf-life, etc. The use-case scenario of a food distribution warehouse serves as a model for automating warehouse operations in many industries where rapid response to changing market needs is required. ILIAD aims to develop automated solutions to the complete range of tasks required for the intra-logistics chain in this type of scenario. However, the expected impact of ILIAD also goes well beyond the warehouse context. ILIAD develops key technologies that are relevant to all kinds of multiple-actor systems where robots and humans operate in the same environment. We expect to extend the state of the art in the fields of robot perception (including reliability-aware mapping and learning of semantic maps), planning (task allocation, coordination, motion planning), navigation, manipulation, and human–robot interaction. All of these innovations are essential for enabling independent, coordinated, safe and reliable operation of robots in shared human–robot environments.
The overarching goal of ILIAD is to address limitations in the state of the art which impede the efficient use of robot fleets in warehouse logistics. We address these limitations by the following means: a systematic study of human safety in shared environments, the development of a generic, safe and efficient solution for a mixed fleet of robots handling logistics tasks in human–robot shared environments, supporting life-long operation (meaning that the system can run independently, even when the environment changes), efficient methods for fleet coordination (of mixed fleets of autonomous and human-driven vehicles), and automated picking and handling of a wide range of goods without replacing the gripper.
In order to drive the proposed research and innovations, and to maximise the impact of these actions within the industry, ILIAD has adopted a particularly demanding use case for logistics in the distribution of food products, involving AGVs operating together with human workers. For ILIAD, the food industry provides an especially relevant use case because of its particularly challenging requirements: sensitive products with short shelf-life, etc. The use-case scenario of a food distribution warehouse serves as a model for automating warehouse operations in many industries where rapid response to changing market needs is required. ILIAD aims to develop automated solutions to the complete range of tasks required for the intra-logistics chain in this type of scenario. However, the expected impact of ILIAD also goes well beyond the warehouse context. ILIAD develops key technologies that are relevant to all kinds of multiple-actor systems where robots and humans operate in the same environment. We expect to extend the state of the art in the fields of robot perception (including reliability-aware mapping and learning of semantic maps), planning (task allocation, coordination, motion planning), navigation, manipulation, and human–robot interaction. All of these innovations are essential for enabling independent, coordinated, safe and reliable operation of robots in shared human–robot environments.
Notable outcomes include methods for easy deployment and reliable operation such as automatic sensor calibration, methods for merging and using prior map info to speed up deployment, as well as flow-aware maps that can be used to learn and predict how things move, in order to learn to better blend in with existing operations.
We have also innovated in reliable people detection and tracking in a wide range of conditions, combining cameras and laser scanners.
We have contributed to human safety, in partular through a unified representation for biomechanics impact data and robot dynamic properties, which is to be used for safety-aware motion planning.
We have also developed novel motion-planning algorithms that improve the efficiency of planning in tight spaces, and approaches that plan motions considering learned human behaviours. Other notable results include safe and deadlock-free coordination also under poor network conditions.
We have also implemented a two-arm object picking system and a corresponding vision system to precisely detect object poses. Finally, we have developed and implemented a cutting tool along with a control and vision system that makes it possible to detect and cut open stretch wrap packaging.
We have also innovated in reliable people detection and tracking in a wide range of conditions, combining cameras and laser scanners.
We have contributed to human safety, in partular through a unified representation for biomechanics impact data and robot dynamic properties, which is to be used for safety-aware motion planning.
We have also developed novel motion-planning algorithms that improve the efficiency of planning in tight spaces, and approaches that plan motions considering learned human behaviours. Other notable results include safe and deadlock-free coordination also under poor network conditions.
We have also implemented a two-arm object picking system and a corresponding vision system to precisely detect object poses. Finally, we have developed and implemented a cutting tool along with a control and vision system that makes it possible to detect and cut open stretch wrap packaging.
ILIAD develops key technologies that are relevant to all kinds of systems where robots and humans operate in the same environment. The technologies researched within ILIAD will have impact in warehouse intralogistics as well as other applications of industrial robotics and automation. Crucial hindrances faced today by the industry in five domains are: dependability, vehicle types, efficiency and safety, the dynamic nature of the environment, and planning. These domains are completely covered by the objectives of the project. Beyond the intralogistics context of the project, we expect a sustainable scientific exploitation by extending the state of the art in the fields of robot perception (including reliability-aware mapping and learning of semantic maps), planning (task allocation, coordination, motion planning), navigation, manipulation, and human–robot interaction in mixed human–robot environments.
ILIAD has contributed to the field of localisation and mapping by improving ease of use and accuracy. Using sensor self-calibration and novel quality metrics, the system is also reliability-aware. We have extended state-of-the-art in long-term operation via learning from past experience, better taking into account temporal context, such as the current time of day.
As for planning and coordination, we have made strides towards more compliant robots by means of human-aware maps and navigation, as well as innovations leading to coordination that seamlessly integrates motion planning, task allocation, and robot controllers.
As for manipulation (picking and handling of objects), ILIAD has designed novel grippers that can handle containers that differ in size, weight, and softness. In addition to handling heterogeneous objects, a further complication arises from the fact that the pallets are wrapped by plastic film. Cutting open this film and correctly picking all the objects on the pallet require delicate and flexible manipulation skills that go well beyond what state-of-the-art systems can handle.
ILIAD has contributed to the field of localisation and mapping by improving ease of use and accuracy. Using sensor self-calibration and novel quality metrics, the system is also reliability-aware. We have extended state-of-the-art in long-term operation via learning from past experience, better taking into account temporal context, such as the current time of day.
As for planning and coordination, we have made strides towards more compliant robots by means of human-aware maps and navigation, as well as innovations leading to coordination that seamlessly integrates motion planning, task allocation, and robot controllers.
As for manipulation (picking and handling of objects), ILIAD has designed novel grippers that can handle containers that differ in size, weight, and softness. In addition to handling heterogeneous objects, a further complication arises from the fact that the pallets are wrapped by plastic film. Cutting open this film and correctly picking all the objects on the pallet require delicate and flexible manipulation skills that go well beyond what state-of-the-art systems can handle.