Autonomous waste collection in your company

A 30% saving on the cost of collecting your company's waste. We want to achieve this by setting up an autonomous waste collection system, together with VIL (Flemish Spearhead Cluster for Logistics), FTSolutions, DSP Automation, Pixelvision, Indaver and the University of Antwerp. As a production or logistics company, you want to be able to focus on your main task: producing or transporting as much as possible at the best possible quality. Unfortunately, most production processes also generate a lot of waste. For instance, Flemish companies produce about 16 million tonnes of primary industrial waste per year. This waste, which is often neatly separated, ends up in various types of containers on the company floor. These containers must then be emptied (usually on a daily basis). After a brainstorm with a number of logistics members, VIL identified a need for automating this process.

Fortunately, there are companies such as Indaver, which provide a service for collecting waste from your company site. So as to be able to optimise this service, they are also looking into further automation applications. Automating waste collection is anything but easy. Apart from the fact that several types of containers need to be lifted and moved autonomously (2-wheel, 4-wheel, pallet-based containers etc.), an autonomous solution also needs to work both indoors and outdoors as full containers often need to be emptied outside on the company premises.

Sensors & data augmentation

Autonomous overall solutions for waste collection do not yet exist, but FTSolutions sees an opportunity in their development. Recently, a certified safety sensor has been introduced for autonomous platforms that can also be used outdoors. This offers opportunities for this type of applications. In addition, AI algorithms are getting smarter by the day, making it easier to recognise containers in difficult weather conditions (such as rain, fog and snow); this is a typical task for Pixelvision. Flanders Make is working on data augmentation techniques to train this kind of AI networks. With data augmentation, new images are created in various weather conditions (Figure 1).

Autorio----Figuur-1Figure 1: Left, a recorded camera image of containers in dry weather. Right, an image created by an algorithm in rainy weather conditions.

Based on these images, an AI network can learn, for instance, what a container looks like in the rain. Together with new ultrasonic technology (Figure 2) developed by the University of Antwerp, we can then integrate a robust solution.

Autorio----Figuur-2Figure 2: The new ultrasonic sensor developed by the University of Antwerp.

We also ensure a comprehensive evaluation of various sensors. By conducting measurement campaigns, we can select the right sensor for every weather condition and map out how to cleverly combine these different sensors to generate enough information from the environment for autonomous waste collection. For example, in certain weather conditions, such as fog, an ultrasonic solution will work much better than a camera or a LIDAR. This evaluation and subsequent ideal combination of sensors allows us to provide the best possible solution at the lowest possible cost.

Ultimately, an autonomous pick-up service must also be integrated into a warehouse management system. In order to ensure that the waste will be collected smoothly, we need data on the containers. This includes the type of container, whether it is in use or empty, the content of the container, etc. By collecting and analysing all these data, we will in the end be able to collect waste much more optimally.

Autonomous mobile platform design

During the start of the research, all requirements were identified; for instance, the autonomous mobile platform will also have to be able to drive on less slippery road surfaces, place a lid on a container, etc. There are various design possibilities for this. Together with FTSolutions and DSP Automation, we’ve analysed which design would probably be the most favourable in terms of cost and robustness.

Proof-of-concept

By developing software combining this sensor information with controllers that lift and transport the various containers, we can create a proof-of-concept. We demonstrate this proof-of-concept on our open mobile research platform. With this mobile platform, we test the sensors and software in all kinds of realistic conditions.

IMG_5521Figure 3: The proof-of-concept AGV with the various types of containers in our test warehouse in Lommel.

You may remember the Disney film Wall-E, where an autonomous robot cleans up the world's rubbish long after humans have left. This somewhat sad picture of the future will hopefully be partly prevented by "our own Wall-E" solution as it contributes to smart, optimal and separated waste collection services now.

In about a year's time, we will demonstrate the final results of this research. We will gladly invite you to this event. Would you like to see our latest innovations for yourself and follow a demonstration? Please fill in the form below and we will contact you.

Ellen van Nunen - Project leader
Auteur

Ellen van Nunen - Project leader

Ellen van Nunen has been working at Flanders Make since 2018 as project leader on projects around autonomous driving. Ellen obtained her Master of Science in Applied Mathematics in 2004 (TU Eindhoven) and has over 10 years of experience in research on autonomous vehicles.

Related Posts

STAY INFORMED!

Sign up for our newsletter and receive the most interesting articles about innovations in the manufacturing industry and updates about Flanders Make.