Skip to content
Autonomous Object identification
Autonomous Object identification View original picture , opens in new tab/window

AutoPlant

Published: 10 February 2021

Autoplant is part of the second stage of the Vinnova-funded research program Challenge-Driven Innovation (Swedish: UDI), where the goal is to solve societal challenges through collaborative projects that contribute to the sustainability goals in Agenda 2030. The main goal includes technical and logistical solutions for reforestation with driverless machines.

Background

The forest is central to the circular bioeconomy. Through photosynthesis, growing forest absorbs carbon dioxide and large amounts of carbon are today stored in forests and forest land. There is therefore a great societal challenge in using the forest without negative environmental impact and to enable the substitution of fossil-based products at the same time as the standing forest is needed as carbon sink. The solution is sustainable forestry methods with high and value-creating forest growth through resource-efficient precision technology. This enables the use of the forest as a bio-based resource at the same time as the total timber and carbon stock can increase.

All forest planting in the Nordic countries today is manual, with the exception of a few planting machines as experiments. In Sweden, some 300–400 million seedlings are planted annually. The work is low paid and is largely performed by foreign seasonal workers. Although the work is heavy, it is a job opportunity that with the project's solutions in place in the future will probably be replaced by fewer and more educated workers. A goal conflict with the project will thus be greater security and quality for Swedish forest growth at the cost of fewer jobs for these seasonal workers. Another goal conflict may be that the reforestation operations give rise to emissions of e.g. carbon dioxide. However, these are considered rather insignificant in relation to the project's climate benefit and in that the project itself entails an increase in efficiency where the total energy consumption is less than with traditional site preparation methods.

Today, there is no complete (autonomous) solution that includes all parts of forest regeneration. There are continuously advancing planting machines, which prepare and set seedlings at regular intervals and are controlled by a machine operator in a cab, but no solution where the machine itself can fill and feed seedlings, choose planting points and avoid obstacles such as stumps and stones. Autoplant project therefore distinguishes itself by choosing an optimal point for site preparation and planting and being completely autonomous.

Goal

The overall goal of the project (for all three VINNOVA UDI steps) is to create an autonomous tree planting machine where innovative technology allows significant improvement of the working environment and rejuvenation results while minimizing environmental and climate impact. To get there at the end of step three, the goal is to produce functional sub-products in this step 2 - in which AutoPlant now subsides:

  1. a tool for system analysis of an autonomous tree planting machine
  2. one or more units for gentle site preparation and precision planting as well as a system for automatic plant feeding where the plants are not damaged
  3. a vision system that can collect data on terrain obstacles and people / animals in the work area
  4. a tool, “Pathfinder”, for planning the driving route on the forest land (including retrieval of new plants) so that already planted areas are not damaged while taking into account soil moisture, slope and areas where special nature and cultural considerations are required.
  5. a tool for selecting planting points using AI
  6. a tool for self-evaluation of the planting design (the plant is straight, reasonably deep, surrounded by mineral soil, etc.).

Furthermore, work is required to integrate these subsystems into a realistic whole for future steps.

Research subject Machine Design

Machine Design has a key competence in some of the areas that are necessary for the development of an autonomous tree planting machine, e.g. machine design, product development methodology, sensor systems, modeling and simulation. In addition, the research group has a unique research infrastructure for testing autonomous machine systems in terrain (a modular machine of 10 tonnes). Machine Design contributes with knowledge of external and internal perception and has experience of research and development of systems for identifying objects in forest environments (rocks, stumps, etc.). Machine Design wants to develop its knowledge of automating various forestry activities in general and of land preparation and afforestation in particular and expects to develop new methods and tools, validate these in the forest environment and publish relevant research articles.

PhD Project 1: Object Detection and Identification

One of the basic preconditions for the unmanned forest machine platform to be able to perform automatic operation, is to achieve the ability to perceive the surrounding environment with the help of the vision system. The system consists of image sensors, processing and computing units, to realize parameter estimation of potential targets and obtain information about targets e.g. position, form, attitude and category.

To build the vision system, it is necessary to create the underlying communication between the imaging sensor and the computing unit and to adopt the corresponding software framework in the computing unit to deploy custom algorithms. Through the use of different image signals and the collocation of different algorithms, the construction of different types of local maps of the unmanned forest machine platform is realized, and the guidance signals for the operation of other parts of the platform are output.

According to the actual needs of different projects, the team members analyzed the corresponding operating environment, combined with the specific operations involved, selected the corresponding imaging sensors and developed autonomous algorithms, and finally relied on the unmanned forest machine platform for the actual deployment and testing of the system.

 

PhD Student: Songyu Li https://www.ltu.se/staff/s/sonliv-1.187314

Supervisor: Håkan Lideskog https://www.ltu.se/staff/h/haklid-1.77103