6 th International Forest Engineering Conference Quenching our thirst for new Knowledge Rotorua, New Zealand, April 16 th - 19 th, 2018

Size: px

Start display at page:

Download "6 th International Forest Engineering Conference Quenching our thirst for new Knowledge Rotorua, New Zealand, April 16 th - 19 th, 2018"

Elizabeth Wilkins
6 years ago
Views:

6 th International Forest Engineering Conference Quenching our thirst for new Knowledge Rotorua, New Zealand, April 16 th - 19 th, 2018 AUTOMATION TECHNOLOGY FOR FORESTRY MACHINES: A VIEW OF PAST,

1 6 th International Forest Engineering Conference Quenching our thirst for new Knowledge Rotorua, New Zealand, April 16 th - 19 th, 2018 AUTOMATION TECHNOLOGY FOR FORESTRY MACHINES: A VIEW OF PAST, CURRENT, AND FUTURE DEVELOPMENTS Pedro La Hera a,b, Omar Mendoza Trejo b, Daniel Ortíz Morales a a Swedish Cluster of Forest Technology, b Department of Forest Technology, Swedish University of Agricultural Sciences, Umeå, Sweden - xavier.lahera@slu.se Introduction The forestry industry continues to be radically altered by technology as it seeks increased productivity and greater efficiency. In many industries, automation and the concept of smart systems are technologies beginning to answer questions about productivity, efficiency and safety. In the forest industry itself, automation is a concept that prompted long ago, around the 1990s, because of all the benefits that are seeing possible with the upcoming of such a technology. But with current advancements in the fields of robotics and artificial intelligence, we ask ourselves: where is the forest industry standing in this area today? And the reason to raise this question is because we observe that for the past decade the engineering of forestry machines has taken an opposite approach to what many would have expected. To understand this statement, let s look at some of the shortcomings observed in today s forestry machines: Lack of modularity, with costly manufacturing, and unpractical adaptation to new applications, including mechanized tree planting, soil preparation, forestry equipment carriers, etc. Heavy mass and long cranes, limiting mobility, representing high fuel costs, and resulting in non-ergonomic levels of residual vibrations that are harmful to the human body. Lack of motion/vision sensing, GPS, and outdated embedded processors, preventing the implementation of modern control software to assist in navigation and manipulation. Inability to communicate to provide information or to realize machine-to-machine collaboration. Complex joint-by-joint and motion-by-motion control, with non-intuitive commands via buttons and joysticks. Such a solution influences user friendliness and learning, but it also makes it unpractical to introduce more advanced multi-tasking machines (for instance, machines with multiple cranes). Need of high pace mental multitasking and experience, factors that affect mental performance and productivity. So, why is it so that an industry that demands greater efficiency than others continues working with 20 th century technology? Could it be that the industry is not interested on newer technology, despite the benefits? Or could it be that automation technology currently has limitations preventing the forest industry from adopting it? 1

2 The simplest answer to these questions is that there is more happening in the area of forestry automation than the eyes can meet. Unfortunately, researchers are no longer making work of this nature publically available due to industrial interest. In parallel, industry is starting to adopt some initial technologies that will lay the ground to automation. But although development is not as opened as we would expect, we can analyze the current situation by using a standard road map of development broadly applied in the field of automation. Combining this road map with a) scientific publications found in the area of forestry robotics, and b) current market products, can give us an estimation of the past, present and future of what we can expect happening in the forest industry. In the text below we limit our focus to the field of machine automation. Other benefits that come with the appearance of sensing technology and automation will fall out of scope. For instance, better data analytics, wireless network communication, or better user interfaces, just to mention some. To facilitate presentation, we will start by defining automation according to international standards. These standards will be useful to relate the state-of-the-art, and also to show a road map showing the past, the present, and future developments in the automation of forestry machines. Levels of automation The degree to which a given task is automated is known in engineering as Level of Automation (LOA) [1]. LOA serves to explain the ability of an algorithm to perform a given automatic function, and how much human involvement there is in the process. Although the terminology of LOA varies slightly, it is quite similar in most fields involving automation technology: robotics, artificial intelligence, and automatic control, to mention a few. Currently, a summary of five levels exists to educate the wider community about the stepwise progression of automation. A comprehensive definition of these levels is presented in [1]. For the sake of simplicity, we shortly give an overview of these LOA in Table 1, a table also providing examples related to the automotive industry, as most people know how a standard vehicle operates. TABLE 1: LEVELS OF AUTOMATION LEVEL DESCRIPTION HUMAN EXAMPLE INVOLVEMENT 0 Operator only A human operator Automotive industry before the year 2000 when the performs all tasks automotive industry started equipping cars with computers and sensors 1 Operator Assistance Basic simplified A human operator performs all tasks, but it receives control functions computer support simplifying some actions. Some examples include automatic transmission, cruise control, or anti-sliding control. 2 Partial Automation Function specific Vehicles performing automatic self-parking, or automation automatic breaking to avoid collisions. 3 Conditional Limited self-driving A vehicle trained to drive on a city, but under constant automation automation supervision of a person. The ability to reason outside a given set of conditions is limited. 4 High Automation Fully automated in A vehicle trained to drive on its own, and it does not a defined use case require supervision of a person, but it will ask when a situation is out of its data base. 5 Driverless Fully automated in A vehicle driving on its own, and it does not require all use cases any supervision, as it is able to take its own decisions and learn from its surroundings 2

3 The description in table 1 allows us to match the state-of-the-art in automation according to different industries. In relation to current reality, the LOA tell us that the highest level of automation currently existing in the most advanced industries is level 3. Some industries showing the abilities of automation-level 3 include the automotive, robotics, military, or aerospace industries, where systems equipped with advanced Artificial Intelligence (AI) and embedded hardware are able to compete against human skills. But while these autonomous systems are good for certain conditions (conditional automation), they are not good at everything, especially tasks like learning, making maps, easily identifying objects, or other basic human abilities needed to accomplish more advanced actions. Reaching level 3 has taken historically 50 years of research, and we are just in recent initial years of adoption worldwide. This shows the bold reality that adopting technology for industry can take more than a decade from the time initial prototyping results are seen at the research level. Current developments in the areas of automation involve efforts to take technology above automation-level 3, but such developments will take years to reach maturity. Examples of research problems include: learning from demonstrations, understanding spoken language, wireless network communication, or quickly identifying objects in images, just to mention some. Experts believe that automation-level 4 will be achieved until the year 2025 in terms of research. Industrial adoption will take more than another decade to complete. Level 5, however, brings us into the more distant future, where autonomous systems may perform themselves in all cases, completely without human supervision. Relating the LOA with ongoing developments in the forest industry Harvesters and forwarders are machines attempting to automate timber-harvesting processes. These machines harvest, stack and move logs in and out of the forests. But several needs motivate the search for technology that will allow the industry to operate more effectively: The need to facilitate operator actions The need to use fuel as sparingly as possible The need for operator safety and comfort at every step of the process The need for better planning methods From Level 0 to level 1 automation A harvester operator sitting for an eight-hour shift faces a number of potential productivitybusters. One example is the complex coordination required to seamlessly produce joystick commands resulting in motions of the crane and vehicle. Precise control of the many crane links and harvester-head usually requires a series of expertly coordinated motions that can prove tiring over time. This is the current situation of automation -level 0, where the forestry industry is largely standing today. One method of automation in this area is the concept of computer-assisted motions, whereby an operator receives help from a computer to perform better motion patterns. This scenario raises an improved partnership between man and machine for operating cranes and can help new to medium operators perform precise control more intuitively and in a simplistic way. 3

4 According to the LOA standards, providing computer-assistance for motion control falls in the category of automation-level 1. One of the challenges to reach this level with heavy duty equipment is the necessity for better hydraulic control systems, which are currently quite rare in the market. The lack of smart hydraulic systems represents a limitation for the adoption of this technology. Nevertheless, we are currently at a point where we see the forest industry finally transitioning towards the beginning of this technology. What we find in the market are entry level examples to automation-level 1, some of which are: Cranes equipped with motion sensors, providing the entry level to improve motion control software [2]. Basic boom-tip control, where the operator receives computer-support to perform expertly coordinated end-effector movements with less effort. Currently, only an entry-level basic method exists on the market [3], but the robotics industry shows the path to several possibilities. For instance, different tip-control algorithms can be designed according to selectable optimization options: minimum kinetic energy control, minimum potential energy control, failure recovery, strength optimization, and fuel consumption. Some of the research results in this area are shown in [4, 5, 6]. Reduced crane vibrations, making the operation of the crane more comfortable [3, 7, 8]. Active suspension, improving the ride quality in uneven terrains [9]. Hydraulic valves equipped with digital electronics, providing the necessary tool for an entry level to improve software for dynamic motion control of the machine, and develop smart hydraulic control systems [10]. The technologies listed above show how current developments are starting to consider the hardware requirements and initial software needed for automation. But transitioning towards this technology will not be easy, because developing software and redesigning all hydraulics and embedded electronics for forestry machines will present difficulties along the way. And we have to keep in mind that companies have to profit along this process. Therefore, entering the world of automation-level 1 will be a difficult step, and it will take the forest industry at least 15 years to complete. But in those years we can expect to already see improvements in control performance, particularly precision boom movements using motion sensors and operator-assistance software. According to the research we find available, creating smart motion for forestry cranes will rely on libraries containing specific automated functions, many of which have already been developed by scientists during the past years [11, 12]. However, the operator will be essential to properly use these features, and many difficult motions will still be performed manually. From Level 1 to level 2 automation At the later stages, we expect that in this emerging human-machine partnership operators will take advantage of advanced capabilities of data through computer vision. Machines will present the possibility to select trees by pointing to their location. Consequently, expertlycoordinated automatic motions will either harvest or collect trees, dramatically increasing productivity and reducing operator fatigue. The appearance of such a technology in the forestry sector will indicate that the industry has reached automation-level 2, where operating 4

intervals of times. Nonetheless, most planning tasks will be performed by people, who will also perform tasks manually in very difficult situations.

5 the machine will be merely the work of automated computer functions performing dynamic motion control. The operator will coordinate the tasks of the machines, and by then, technology will enable the possibility to operate many cranes simultaneously, because cranes will operate autonomously for short intervals of times. Nonetheless, most planning tasks will be performed by people, who will also perform tasks manually in very difficult situations. For both of these cases, the user interfaces will become simpler, because many unnecessary buttons and joysticks will be replaced by software algorithms. Research developments targeting these goals have already reached a high level of maturity [6]. Figure 1. Technology will enable the possibility to operate many cranes simultaneously, because cranes will operate autonomously for short intervals of times [13]. Figure 2. Machines will present the possibility to select trees by pointing to their location. Consequently, expertly-coordinated automatic motions will either harvest or collect trees, dramatically increasing productivity and reducing operator fatigue [14]. 5

6 From Level 3 and beyond Reducing energy consumption is an important target in the forestry industry. Typically this involves looking at the total system and asking how it can deliver the most output from the given fuel. Given continuous harvester or forwarder operation - all year round - small gains in fuel efficiency have an enormous impact. However, forestry machines are becoming larger and heavier, increasing their cost, noise/vibration levels, control complexity, and decreasing their mobility. As a consequence, forestry machines are lacking energy efficiency, and they are costly to produce. Level 1-2 automation will certainly increase the energy efficiency in machines, but they also have limitations in this aspect. From automation-level 2 further increases in efficiency will point out to the overall design of the machine. Leveraging fuel efficient systems will constitute a new, but interesting challenge, because it will enable the possibility to rethink how machines are designed from the bottom ground. Passing automation-level 3 will show that human operators will not necessarily need to sit on a cabin. Thus, designing smaller and lighter machines will be opened to possibility. Having machines without people will reduce the need for comfort and ergonomics. Therefore, machines will be cheaper to manufacture, and most of the costs will constitute the hardware, software, amount of cranes/tools, and power generation existing by then. Machines of this kind will present better motion capabilities, better power sources, and dynamic motion control, all of which will contribute to the overall energy efficiency. In later stages, machines will start presenting bio-inspired designs, because designs of this kind will improve efficiency in off-road navigation, and gentleness towards the soil [20]. Figure 3. Having machines without people will reduce the need for comfort and ergonomics. Therefore, machines will be cheaper to manufacture, and most of the costs will constitute the hardware, software, amount of cranes, and power generation existing by then [15] 6

Figure 4. In later stages, machines will start presenting bio-inspired designs, because designs of this kind will improve efficiency in off-road navigation, and gentleness towards the soil [16].

7 Figure 4. In later stages, machines will start presenting bio-inspired designs, because designs of this kind will improve efficiency in off-road navigation, and gentleness towards the soil [16]. This technological progress will enable a re-structuring of the forestry process, because it will open the possibility to run forest operations almost without people. In essence, this technological steps will open the doors to automate forest operation as an entire process, where machines are seen as single units performing a given task from a chain of operations. Operators will initially sit in a command center nearby the machines, but eventually they will be moved far away, close to cities. Past and current developments have considered these scenarios. For instance, machines without cabins were presented in [17, 18], machines with efficient power sources were presented in [20], and bio-inspired designs were presented in [19], and they are undergoing serious research. Unfortunately, our current level of automation technology limits using these developments in commercial products due to costs and safety concerns. Nonetheless, the main nature of designing systems of these kinds already shows the interests behind these concepts. Conclusions Despite what is commonly believed, many research efforts have been put to develop smart automation technology for forestry machines over the past decades. Today, unfortunately, it has become quite difficult to access these (once opened) research projects because of the industrial interest behind them, and because they are starting to become a new form of industry in this area. Nevertheless, this extended abstract has tried to shortly show some of the developments that are recently being adopted in the market. Also, it has tried to point out some articles showing some of the earliest developments that readers can find available today. As developers of this technology, we find unfortunate that not all development devoted to this area can be openly discussed, because this leaves a big gap of information between what already exists, and what should be the focus of research for new comers. Therefore, as authors we agreed that defining the LOA table, and pointing out some of the products that we can find openly available, will give some researchers out there the ability to see an entry point according to the road map of development in this field. 7

8 Bibliography [1] Institute of Electrical and Electronics Engineers., & IEEE Systems, Man, and Cybernetics Society A Model for Types and Levels of Human Interaction with Automation. IEEE transactions on systems, man, and cybernetics: A publication of the IEEE Systems, Man, and Cybernetics Society. New York, NY: Institute of Electrical and Electronics Engineers. Vol. 30. No. 3. [2] Cranab & Slagkraft. The first fully sensorized forwarder crane. Retrieved from: [3] John Deere. Smooth Boom Control and Intelligent Boom Control. Retrieved from: [4] La Hera, P. (2011). Underactuated Mechanical Systems: Contributions to trajectory planning, analysis, and control. Doctoral dissertation thesis. Umeå University, Sweden. [5] Ortiz Morales, D. (2015). Virtual Holonomic Constraints: from academic to industrial applications. Doctoral dissertation thesis. Umeå University, Sweden. [6] Ortiz Morales, D. (2015). Westerberg, S. (2014). Semi-automating forestry machines: motion planning, system integration, and human-machine interaction. Doctoral dissertation thesis. Umeå University, Sweden. [7] Hera, P. L., & Morales, D. O. (2015). Model-Based development of control systems for forestry cranes. Journal of Control Science and Engineering, 2015, 27. DOI: /2015/ [8] Komatsu Forest. Smart Flow Control. Forwarders/Forwarder-Options [9] Ponsse. Cabin Suspension System for 8-wheelers. Retrieved from: [10] Mathworks, INCOVA Designs Intelligent Valve-Control System for a 20-Ton Excavator. Retrieved from [11] Ortiz Morales, D., Westerberg, S., La Hera, P. X., Mettin, U., Freidovich, L., & Shiriaev, A. S. (2014). Increasing the level of automation in the forestry logging process with crane trajectory planning and control. Journal of Field Robotics, 31(3), [12] Pedro La Hera and Daniel Ortíz Morales Model-Based development of control systems for forestry cranes. J. Control Sci. Eng. 2015, Article 27 (January 2015), 10 pages. DOI: [13] Mellberg, A. (2013). Extension Operator Environment for ForestHarvesters. Master thesis. School of Engineering, Jönköping, Sweden. [14] Luu, T., Sabir, E., Sundelin, L. Forestry in Operating the Harvester Crane. Umeå Institute of design. Umeå University. Sweden. [15] Markus Leijon. Forestry machines in Umeå Institute of design. Umeå University. Sweden. [16] Ludwign, Forestry in Autonomous Harvester. Umeå Institute of design. Umeå University. Sweden. 8

9 [17] Direct Industry. Forestry forwarder Pully. Retrieved from [18] Unmanned Forest Machine. Retrieved from [19] Timberjack Walking Forest Machine Concept by Plustech Oy. Retrieved from: [20] ElForest Technologies. Reference projects. Retrieved from: 9

FORESTRY IN 2020 OPERATING THE HARVESTER CRANE

FORESTRY IN 2020 OPERATING THE HARVESTER CRANE 5ID076 - Project 1: Professional Users MFA Interaction Design, Umeå Institute of Design dh.umu.se/interaction 16.01.2015 Trieuvy Luu, trieuvy@trieuvyluu.nl