Navigation and Tree Measurement
long run. The burden of the operator could be lightened by
increasing the automation level of the forest machine. This
could mean, for example, that the control of the crane
would be partly automated. The operator could just indicate
which tree to cut and the crane would automatically grasp
it. This would give the operator the chance to better utilize
his greatest strength, which is the expertise about the trees.
Increased automation would also make the job easier and
the training period could be shortened for new operators.
The Forestrix project studies advanced sensor systems
for forest harvesters. One of the sensors used is a scanning
laser range finder (henceforth, a laser scanner). As the
machine is moving in the forest, the laser scanner can take
multiple scans of the trees it passes from different directions.
In the Forestrix project, the laser scanner is used for
tree measurement and mapping, and simultaneous
localization and mapping (SLAM). In robotics, SLAM is used
for solving a chicken and egg problem: When the robot
moves in an unknown environment, it has to create a map
of the environment on the one hand, and localize itself
within the same map on the other.
The standard practice for the height of the tree
diameter measurement is equal to 1.3 meters above the
ground. When having only one static 2D laser scanner
onboard, it is impossible to determine the measurement
height without any kind of terrain model. Such a model
was not available at the time and therefore we had to
settle for an approximation with the measurement height.
In the mapping, a GPS receiver was tried for localizing
the measurement platform. However, the thick foliage of
the forest prevents a continuous GPS fix and thus pushed us
to use some other method for localization. SLAM methods
normally require good landmarks in the environment and, in
this case, the trees in the forest are ideal for that purpose.
FIGURE 2. Honda ATV
used in testing the system.
for a 24V electric system utilized by the measurement
devices. An inverter is included for providing 230 VAC
for the PC which performs the data acquisition with a
conventional Windows XP operating system. The actual
data acquisition software is programmed with C++
language. A brute user interface of the program is used to
control measurement devices and data storing process.
Obviously, it would have been too expensive to have a
genuine harvester laying in the backyard of our laboratory
during the whole project (Figure 1). For that reason, we
equipped a Honda ATV (Figure 2) with the necessary sensors
and other equipment for the various stages of the project.
The requirements for the platform in this particular case
were an adequate ability to carry measurement equipment
and the capability to maintain a steady and low locomotion velocity since the forest terrain is uneven. Plus,
vertical smoothness while moving was also required.
We used a SICK 2D laser range finder as the
primary sensor. We mounted the scanner in front of the
measurement platform. It is used continuously at a
measurement rate of 38 Hz. Additionally, a 3D laser
range finder is used, but only for reference measurements and only when the measurement platform is in a
stationary state. On the platform, a data acquisition PC
is also installed. The measurement data are collected
through serial connection and stored in numerical form
to files. The measurement platform includes batteries
The tree map is created by using the SLAM algorithm.
We’ll discuss the operation of the tree measurement
algorithm shortly. They work independently but the tree
measurement information is added to the map in every
cycle of the algorithms. The overall structure of the
feature-based SLAM algorithm is shown in Figure 3.
FIGURE 3. The structure of
the Forestrix SLAM algorithm.