Mobile follower robot as an assistive device for home oxygen therapy – evaluation of tether control algorithms
© Endo et al.; licensee Springer. 2015
Received: 8 October 2014
Accepted: 6 November 2014
Published: 15 February 2015
Home Oxygen Therapy (H.O.T.) is a medical treatment for severe lung diseases in which the patients are supplied concentrated oxygen. This paper investigates the use of a follower robot as a support device for H.O.T. patients, consisting of a two-wheeled differential drive robot connected to the user by tether. Two different control algorithms were studied using dynamic simulation and motion capture experiments with healthy subjects. In further experiments with H.O.T. patients, including a questionnaire survey, it was confirmed that Follow the Leader control was capable of following the user’s trajectory more accurately than Pseudo-Joystick control, and that overall H.O.T. patients showed a preference for Follow the Leader control.
KeywordsHome oxygen therapy Leader following Mobile robot Tether
Chronic Obstructive Pulmonary Disease (COPD) is a common respiratory condition where airflow through the lungs is restricted, often involving permanent lung damage, with patients experiencing coughing, wheezing, and shortness of breath. COPD is an umbrella term, including emphysema and chronic bronchitis, and is usually caused by tobacco smoking (though it can also be caused by exposure to other airborne irritants or pollutants). The World Health Organization reports that COPD is responsible for over 3 million deaths each year, making it the fourth most common cause of death globally . The effect on quality of life can be significant: those with severe shortness of breath may be unable to move around without aid, they may be unable to participate in physical activities, and they may suffer from anxiety and depression as a result [2,3].
The administration of concentrated oxygen for extended periods (over 15 hours per day) can benefit patients with COPD: Home Oxygen Therapy (H.O.T.) aims to further improve the patients’ freedom and quality of life by allowing treatment outside of hospital, and previous research has shown a positive correlation between average daily distance walked and health related quality of life . There are currently around 150 000 people using H.O.T. in Japan, and this number is expected to increase as Japan’s population ages in future. Oxygen is delivered through a mask worn on the face or nose, through a cannula, from a supply which usually consists of either a canister of pressurized oxygen or a liquid oxygen tank. This equipment typically weighs around 4 kg, and when the user leaves the house they can use a small handcart to transport it. Despite the benefits of H.O.T., it still imposes considerable restrictions on the users’ movement and quality of life, since they must expend valuable effort to carry or pull the H.O.T. equipment.
Capable of carrying the H.O.T. equipment (oxygen tank or concentrator)
Capable of following the patient’s movement in daily life
Simple to use
Low weight and compact size
Since H.O.T. requires the use of a cannula to supply oxygen, the system is inherently tethered and this represents a good opportunity to use a tethered robot follower. Tethers — flexible cord-like members with tensile strength but low (or zero) compressive strength — have been widely used in robotics as they are robust and low-cost; they provide a means of mechanical support and leader tracking between vehicles; and they can also facilitate power sharing and communication .
Tethered control methods have been developed previously which allow a mobile robot to follow a leader, using a winch to measure the length and orientation of a tether connected between the leader and the follower robot . Although follower robots have proven successful using the previously developed control methods in person following experiments, they have not been tested with H.O.T. patients. Testing with patients is crucial to evaluate the suitability of a robot and control system, as their needs may differ significantly from healthy users, and a number of user factors may affect suitability, including gait, walking speed, reaction to obstacles and the user’s perceived effort while operating a robot.
We describe the two leader following control algorithms and evaluate them using computer simulation and motion capture hardware experiments in a controlled environment. We assess the suitability of each algorithm for use in an assistive follower robot.
We present the results of experiments conducted with H.O.T. patients, including a questionnaire to assess the needs of the users and evaluate the robot’s performance.
For a simple, low-cost, reliable mobile platform which is capable of following a leader we have proposed using a two-wheeled differentially steered robot . Two control algorithms for such a robot are described below.
In Equation (6), matrix A transforms the wheel angular velocities to body velocities, where r is the wheel radius.
Follow-the-leader control with constant distance
If the robot can determine its own posture relative to an inertial reference frame Σ g X g Y g , more sophisticated tracking of the leader position can be achieved by recording the position of the tether tip over time (see Figure 1b). We define the trajectory of the tether tip as T(s t ) and the trajectory of the follower robot as P(s r ), where s t and s r represent the distance travelled along each respective trajectory. We can calculate the target angle using the following steps:
Θ is the tether angle in the inertial reference frame Σ g X g Y g .
Constant distance control
At this point, the leader trajectory is known, so it is possible to select some forward point on this trajectory and command the robot to steer towards it. However, while the robot should accurately converge on the leader trajectory, there is another important consideration for this application. H.O.T. uses a cannula of limited length between the oxygen supply on the robot and the user, so to avoid stressing this cannula it is necessary to keep the distance between the robot and the user constant (or close to constant).
V r and Ω r can then be transformed into desired wheel velocities (ω L ,ω R ) using the inverse matrix A −1 (6). In this paper, we refer to this second algorithm as Follow the Leader Control for simplicity.
To investigate the performance of the follower robot, we developed a dynamic simulation with the open source V-REP software package, using the Bullet physics engine. V-REP was selected because it allows different experimental conditions to be modelled in a relatively short time, and it has been widely used for a range of robotics application . For this research, we modelled a two-wheeled, differentially steered follower robot (similar to Figure 1) with a frictionless caster at the front and at the rear. The leader was modelled as dummy point moving along a predetermined path at a fixed speed of 0.5 m/s. A model tether connected the robot to the leader, and a sensor provided length, l m , and angle, θ, data to be used in the control algorithm.
In order to characterize and compare the two control methods detailed in the previous section, both Pseudo Joystick and Follow the Leader algorithms were implemented in the V-REP simulation. The robot was first set to use Pseudo Joystick control to follow the dummy leader as it moved along the pre-set path, and the resulting trajectories were plotted. The simulation was then repeated using Follow the Leader control with the same leader path.
When using Follow the Leader control, the robot’s trajectory follows the leader trajectory much more closely (Figure 2b), though we still see some small deviation from the leader path when the robot moves around curves. It is possible to reduce this deviation by increasing the gain K b , however this can decrease stability and lead to dangerous over-rotation (risking damage to the oxygen cannula). To quantitatively compare follower performance, we have used normal path deviation as a metric: we divide the leader trajectory into small segments, and calculate the normal distance to the robot trajectory for each segment. Figure 2c compares the two algorithms using this metric, and clearly shows that Pseudo Joystick exhibits greater deviation from the leader path.
Experiments in controlled environment
Following the simulation, we conducted experiments to validate the follower performance in a hardware prototype.
Follower robot specification
Dimensions L × W × H
670 × 330 × 350
Max. Step Height
Follower performance in motion capture experiment
Normal deviation from leader path (m)
Follow the leader
Experiments involving H.O.T. patients
After confirming that the robot could perform adequately in a controlled environment, further experiments were carried out with H.O.T. patients to assess the robot’s suitability as an assistive device for Home Oxygen Therapy. We conducted a simple follower experiment (similar to the motion capture experiment described above), and used a questionnaire to gather patient feedback about the robot’s performance.
We conducted our evaluation in January 2013 at a Meeting for the Pulmonary Rehabilitation Studies in Nagano, where 14 people volunteered to take part in a practical robot experiment and a questionnaire survey. We obtained informed consent from all the participants before starting the experiments, and no form of compensation was given. This evaluation with H.O.T. patients was approved by the ethical review board for epidemiological study in the Tokyo Institute of Technology (approval No. 2012014).
Position tracking from video data
Due to the limited space available in the testing area, and the risk of burdening patients by attaching sensors, it was not possible to use motion capture or other sensing equipment to record the position of the user and the robot in real-time. For this reason, the experiments were recorded with a video camera, and this video data was later analysed to determine the trajectory data. The video was analysed, frame by frame, to record the position of the user’s feet when they struck the floor. This position was then compared to a known map of the experiment floor to measure the position, and the positions of the left and right feet were averaged to approximate the user’s center of gravity (motion capture experiments have confirmed that this gives a reasonable approximation of the user’s center of gravity). A similar procedure was used for the robot’s wheels. Though coarse, this procedure allowed rough trajectory tracking without overly burdening patients; we estimate the accuracy to be around ±40 mm. The trajectory data of five subjects, selected at random, is presented in this paper.
Results and discussion
Follower performance in experiment with H.O.T. patients
Normal deviation from leader path (m)
Follow the leader
The responses to Question 1, ‘How easy was it to walk around the cones without colliding with them?’ established a baseline for the effectiveness of the robot in this task (Figure 8). Most of the patients responded ‘Easy’ or ‘Very Easy’, with only one responding ‘Difficult’. This is important as the cones walking task is an approximation of some of the daily activities that real H.O.T. users undertake, such as walking to the shops while avoiding other people, and any assistive device should be able to complete this activity without causing difficulty.
Question 2, ‘Which control method was better: A (Pseudo Joystick) or B (Follow the Leader)?’, gives a qualitative comparison of the control methods (Figure 8). The results are mixed but show a slight preference for Follow the Leader (8 positive responses) over Pseudo Joystick (4 positive responses). The preference for Follow the Leader may be due to the relative comfort: there is no need to glance backwards at the robot when using it. The fact that other users preferred Pseudo-Joystick control may be explained by the better responsiveness: with Pseudo-Joystick control, the robot will respond almost immediately to a steering input, while Follow the Leader inherently involves a delayed steering response since it records the history of the leader’s position. Thus some users will find Pseudo-Joystick more intuitive in this sense. In addition to evaluating the technical efficacy of control methods it is essential to also consider the users’ preferences; and the fact that different users prefer different control methods may suggest that user-switchable control could improve the robot’s usability.
Question 3, ‘Did you feel any discomfort using A (Pseudo Joystick) or B (Follow the Leader)?’, was asked to identify further usability problems (Figure 8). Among the responses there was a clear trend that Pseudo Joystick was more uncomfortable to use (6 ‘uncomfortable’ responses) than Follow the Leader (2 ‘uncomfortable’ responses). Pseudo-Joystick may be more awkward to use since it requires the user to occasionally glance backwards, and while this is an easy task for a young, healthy user, it is important to note that it places relatively more physical strain on an elderly person (particularly a person using H.O.T.). These results guide further design revisions as avoiding discomfort is of paramount importance in this application: as the goal is to increase the users’ freedom and well-being, the assistive robot must avoid causing any unnecessary distress which could have a negative effect on breathing and overall health. The responses collected so far indicate that Follow the Leader is likely to be a safer choice for H.O.T. users.
This paper investigated the use of a leader following robot as an assistive device for Home Oxygen Therapy patients. We examined and compared two different control algorithms for the robot using dynamic simulation, motion capture experiments in a controlled environment and most importantly experiments involving H.O.T. patients. From the practical experiments we showed that the Follow the Leader algorithm was capable of following the user more accurately than Pseudo-Joystick, but both algorithms gave reasonable following performance in the walking-around-cones task. The questionnaire survey of H.O.T. users identified that overall they found Follow the Leader to be better and found Pseudo-Joystick control to be more uncomfortable. Pseudo-Joystick control is likely to be more intuitive to some users because of its immediate response to user commands, but the need to look back and check the robot’s position can introduce some discomfort. In addition to the data and analysis presented in the paper, it is important to highlight the active role of H.O.T. users during experiments. It is essential to involve such stakeholders early in the design process, and the members of Meeting for the Pulmonary Rehabilitation Studies in Hokushin were enthusiastic about helping in the development of assistive robotics, providing the researchers with invaluable feedback and encouragement.
The authors would like to express our deepest gratitude to the members and staff of Meeting for the Pulmonary Rehabilitation Studies in Hokushin and Hokushin Flying Disc Club. The questionnaire-based evaluation was also financially supported by the Association for Technical Aids (ATA).
- World Health OrganizationTop 10 Causes of Death: Fact Sheet No. 310. http://www.who.int/mediacentre/factsheets/fs310/en/.
- Kida K (2011) Home oxygen therapy in Japan: clinical application and considerations for practical implementation. Jpn Med Assoc J 138(12): 99–104.Google Scholar
- Pauwels RA, Buist AS, Calverley PM, Jenkins CR, Hurd SS (2012) Global strategy for the diagnosis, management, and prevention of chronic obstructive pulmonary disease. Am J Respir Crit Care Med 163(5): 1256–1276.View ArticleGoogle Scholar
- Janssens JP, Rochat T, Frey JG, Dousse N, Pichard C, Tschopp JM (1997) Health-related quality of life in patients under long-term oxygen therapy: a home-based descriptive study. Respir Med 91(10): 592–602.View ArticleGoogle Scholar
- Krishna M, Bares J, Mutschler E (1997) Tethering system design for dante ii In: Robotics and Automation, Albuquerque, New Mexico, 1997, vol. 2, 1100–1105.. IEEE, New Jersey, USA.Google Scholar
- Fukushima EF, Kitamura N, Hirose S (2000) A new flexible component for field robotic systems In: Proceedings of the IEEE International Conference on Robotics and Automation, San Francisco, California, 2000, vol. 3, 2583–2588.. IEEE, New Jersey, USA.Google Scholar
- Tani A, Endo G, Fukushima EF, Hirose S, Iribe M, Takubo T (2011) Study on a practical robotic follower to support home oxygen therapy patients - development and control of a mobile platform In: IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2011, 2423–2429.. IEEE, New Jersey, USA.Google Scholar
- Rohmer E, Singh SPN, Freese M (2013) V-rep: A versatile and scalable robot simulation framework In: Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 1321–1326.. IEEE, New Jersey, USA.Google Scholar
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited.