Although the use of automated vehicles is now an everyday feature of many sectors of the off-highway industry, the regulatory framework for such activities is still patchy at best. In an attempt to address this lack of regulatory clarity, the Transport Research Laboratory, with support from Innovate UK and the Centre for Connected and Automated Vehicles, has devised an overarching code of practice for the operation of all types of off-highway vehicles. In this presentation, Dr Ianto Guy, lead author of the Off-Highway Automated Vehicles Code of Practice, will discuss how this document can be used to ease the process of adopting automated vehicle technology and the next steps required to support the growing automated off-highway vehicle market.

The agricultural industry is facing serious challenges today and in the future. Precision agriculture, fewer employees and heavier machines will require new solutions. Established agricultural engineering companies are expanding their portfolios to include autonomous machines and robots, while new suppliers and startups are breaking new ground. Companies like Horsch illustrate the directions that autonomous agricultural machinery will take in the future. This presentation is based on research on autonomous agricultural machines and robots, analyzing these machines’ situations and future development. These new categories of equipment will have a significant influence on suppliers.

Computer vision technology is advanced and widely used in automotive industries both for road driver’s assistance and autonomous driving. However, its application on industrial vehicles is not popular. We are now extending and adapting the lessons and developments in automotive to industrial vehicles. Industrial vehicles are bulky and usually operate in a very tough and noisy environment. Furthermore, most of them operate autonomously. These factors create operational issues relating to safety where these vehicles operate. Using computer vision technologies, sensors such as cameras, lidars, radars and sonars can support the drivers of these vehicles, with a 360° view, warning them of imminent danger and taking control of the vehicle in some cases.

In Norway, construction machines are idling for almost half of their working hours. In addition, they are responsible for one-fifth of the greenhouse gases from the building and construction industry. Instead, imagine a road construction project where each machine always knows where the others are, what they are doing and what is the optimal way to organize the work. To realize this, Skanska, SINTEF, Ditio and Volvo are collaborating to develop algorithms that can dynamically map the construction site and activity, identify suboptimal working patterns, recognize effective routing, find out which machines are needed where, and coordinate all the machinery.

This presentation proposes ways to evolve from a driver-centric industrial vehicle to automated driving. Special consideration will be given to increasing the level of automation step by step in a way that work results of previous steps can be utilized for later, higher levels of automation. Furthermore, functional safety aspects will be covered in theory as well as on a real-life autonomous vehicle.

Modern and future mobile machines will include several new functions that need to seamlessly work together. In addition to traditional control systems for hydraulics, for example, modern machines will include hardware and software for telematics and autonomous and assistance systems. These all need a slightly different approach and skills in developing subsystems for mobile machines, but they are still heavily interconnected. We will present how to seamlessly integrate these subsystems with Epec’s offering and know-how by using perception-planning-control model approach.

Assistance functions are the essential first step toward automated operation of mobile machinery. Such semi-automation features can significantly simplify machine usage even for inexperienced operators. They can also improve productivity and contribute to safer construction sites. Combining system expertise in hydraulics, sensors and controls, Bosch Rexroth has developed features such as virtual walls, grading assistants and payload estimation for a broad range of machinery. In this talk, we give an overview of our advanced assistance functions for loader and excavator applications.

World-scale environment simulation has become an essential tool in developing autonomous systems. The automotive industry has proved the effectiveness of employing real-time synthetic environments to train and test AI/machine learning for perception, planning and control. This approach has been adapted to the off-highway sector by integrating the use of real-time equipment simulation and advanced visualization of interactive real worlds with game engines. The presentation will discuss the challenges of deploying and integrating simulation and sensor visualization together with the autonomy software stack to test and train operations in unstructured off-highway environments. We will present how virtual environments can help address the challenges of autonomy and how they can be rapidly deployed in the product development process.

Agricultural machines operate across a wide variety of use cases as they execute their tasks. Even machines that are focused on specific tasks will need to change their operation mode based on conditions in their core operation and for secondary purposes such as movement between work areas. In operator-driven equipment, it is the operator that makes all of the small decisions needed to adapt the machine operation based on these conditions. However, for autonomous machines, a different approach is required. The complexity of adapting to the changes in operations for autonomous machines is often underestimated, and managing these effectively can cause difficulties in system development, resulting in poor performance and reliability. Borrowing concepts from robotics applications, new approaches can be applied to adapting between many complex use cases while maintaining modularity and enabling improved verification and validation testing, providing a solid framework on which to build scalable autonomous machines.

Autonomous vehicles have become more commonplace in mining as operators discover the key benefits of leveraging the technology across their operations. ASI Mining works with mining companies to help realize substantial gains in safety, utilization of assets, productivity and consistency of operation across multiple mining applications. This presentation will share practical insights into how value is extracted from the technology, critical lessons learned in the application and operation, and other considerations that are essential to achieve and sustain long-term gains.

This presentation will focus on the Azure services (IoT, HPC, big data and AI) needed to deploy an end-to-end ADAS/AD solution for the mining and agriculture industries. We will focus on the lessons learned from deploying ADAS/AD at car makers. We will review the business and technology benefits of these use cases.

The trend is clear: off-highway is next in line after automotive to be defined by software. A multitude of highly specialized operations are in the process of being digitalized and automated to increase efficiency and throughput. In this challenge, TTControl sees a key role in software architectures. To boost time and data efficiency, IP and knowledge generation must constantly flow from early proofs of concept until final industrialization, with short iteration loops. Modularization and interoperability with third-party tools are mandatory design goals to cope with a technological landscape in constant evolution and to support a data-driven development strategy.

Driver assistance systems that support machine operators and enable advanced safety for heavy equipment are increasingly based on systems consisting of cameras, radar and lidars. For test cases, petabytes of raw data form the essential basis for function development with AI-based logic. At the same time, they represent a major hurdle in the development process for real-world testing, since the data has to be valuable for training neural networks. The presentation deals with the efficient acquisition and selection of valuable data for automated driving and thus separates the wheat from the chaff.

As the rapid development of sensors and AI algorithms continues to race on, new use cases, methods and industry trends for working with training data are emerging and evolving. This talk describes emerging trends and the use of training data in autonomous mobility applications such as mining, automotive and construction. It also gives insights into how these trends improve performance and create new use cases for perception systems.

Disputes between road authorities and road construction companies often happen around quality numbers, mostly without underlying cause, let alone remedy information. Besides a GNSS annotation on road data, detailed geometry data can enhance the data interpretation and make it uniform for all stakeholders. New automotive lidar technology has been developed to build a detailed digital twin of the road, providing this intuitive data interpretation, including new GIS tools to visualize in the cloud. This presentation describes this lidar trend in road construction, for off-line quality control and interpretation and for online control while constructing the road. The evolution fits an understanding of technologies, merging the automotive and road construction sectors and working on safety and comfort for all road users.

Traditionally, GNSS has been used as the primary localization solution for autonomous vehicles. However, its limitations and unreliability in certain conditions prevent operators from taking the next steps in safety and productivity.  Navtech Radar and Oxbotica have been working together for seven years to solve this challenge. Radar is trusted to perform due to its unrivaled availability and reliable operation in the harshest conditions, overcoming the limitations of other technologies such as lidar and cameras. This presentation explores the challenges and lessons learned in delivering a robust, reliable solution that unlocks the full potential of automation.

Working with technically limited sensor models of the (intended) end product can restrict the potential value of what is achieved through simulation. rFpro’s collaboration with a world-leading sensor developer has helped expose a set of technologies that, together, strengthen and validate the simulation offering. And it’s these rFpro advancements, coupled with far more accurate sensor models from developers, that have led to a shift in perceptions around the use of simulation for the generation of training data. Join Matt Daley to learn more about what’s been achieved, the hurdles we overcame and how the results are progressing sensor development and training data within the industry.

Situational awareness is important not only for fully autonomous solutions but also for improving the safety and intelligence of manned vehicles. With increasing performance and decreasing costs, lidar sensors have the potential to change how machines perceive and interact with the world. However, lidar isn’t perfect. Cameras and other sensors, like radar, have capabilities that lidar can’t deliver. This presentation will dive into the cutting-edge techniques that can be used to fuse the capabilities of multiple sensors with 3D lidar data to provide more complete situational awareness.