What strategies do students use when they are programming a robot to follow a curved line?

In a relatively short period of time, programming has become content in the compulsory school curricula of many Western countries (Balanskat & Engelhardt, 2015; Raptopoulou, 2021). For teachers, this change has been challenging due to for example lack of formal and non-formal experiences in programming, time for transforming the content of the curriculum into teaching, and a great variation in opportunities for in-service training (Vinnervik, 2022a). The present practice-based study focuses on a programming task that was designed and implemented in two technology classrooms in lower secondary schools in Sweden. The study, conducted in collaboration between researchers and two technology teachers, presents and discusses findings showing how students use different strategies for programming a robot to follow a line. Our study, and the analytical procedures used to capture learning as it is constituted in classroom action, comply with the call for studies investigating consequences of teaching for student learning in technology more closely (Hallström, 2018).

Since 2018, programming is part of the curriculum of the Swedish compulsory school. The implementation was preceded by an extensive discussion among various stakeholders arguing for its potential role in the Swedish educational system (Raptopoulou, 2021). Programming did not become a separate school subject but is now stated in the introductory chapters of the national curriculum as essential in the digital infrastructure of today’s society (Swedish National Agency for Education, 2018). Knowledge about programming and its role in everyday life is contextualized as a form of literacy that students should develop through the teaching in every school subject. Programming was also implemented in the mathematic and technology subjects and in the technology syllabus, programming is described as a tool that the students should learn to use for controlling objects. As in many other countries, the technology subject in the Swedish compulsory school has emerged from several different traditions, focusing on technology content in relation to workshop technology, civic education, technology's social consequences and applications in the natural sciences (Hallström et al., 2014; Jones et al., 2013). In relation to this background, it is hardly surprising that there is disagreement among policymakers, teachers, and educational researchers about the content and character of the subject (Danielsson et al., 2018; Fahrman et al., 2020; Hallström et al., 2014; Norström, 2014) and what role programming may play in the technology subject (Raptopoulou, 2021; Vinnervik, 2022a). Teaching and learning programming in relation to the Swedish compulsory school technology subject is therefore little explored and there is yet little experience among the technology teachers concerning the use of programming when teaching technology (Vinnervik, 2022a).

In Sweden, as in many other countries, a simple robot is often the object that the students first meet and potentially will learn to control through programming (Anderhag et al., 2021; Misirli & Komis, 2014). Several studies have described programmable robots as suitable tools for introducing programming concepts but also for developing various skills (Green et al., 2018; Komm et al., 2020; Misirli & Komis, 2014; Xia & Zhong, 2018). The use of robots is often motivated by the fact that programming is difficult to learn (Misirli & Komis, 2014; Sáez-Lopéz et al., 2016) where especially syntax and the logic required to formulate a solution to a specific problem is difficult for beginners (Robins et al., 2003). Also, the concrete and sometimes playful context is considered to be motivating and thus able to support students' learning (Highfield, 2010; Komm et al., 2020; Newhouse et al., 2017). In addition to programming skills and learning specific programming concepts, programmable robots are suggested to support students' social and generic skills such as collaboration competencies and logical thinking (Bers & Horn, 2010). Studies have also shown that programming robots can be important for students' understanding of technology in society and thus also for their technical literacy (Komm et al., 2020; Kurebayashi et al., 2007). Research on younger children's (6–8 years) learning has been of particular interest to the relatively well-known Beebot and Bluebot (Misirli & Komis, 2014) and Beraza et al. (2010) argue for their usefulness, although they also conclude that there are limitations in how they can support a more advanced understanding of programming. Beebots have only five commands, which is a limit to the type of tasks they can be programmed to perform (Beraza et al., 2010; Kazakoff et al., 2013). In a previous study, we examined primary school students’ perception of functionality in their spontaneous programming language for steering a Blue bot. The findings showed that the students primarily perceived a code’s functionality as a question of readability, rather than how well it fit the purpose of controlling the robot, which we suggested—in line with Beraza et al. (2010)—may be connected to the limitations of the programming language that can be formulated (Anderhag et al., 2021). Using more advanced robots with sensors opens up for more complex programming activities by using for example input—output feedback processes. Komm et al., (2020, pp. 259), using Lego EV3 robots, have also shown that such physical—digital systems “can help the students form a viable model of the machine to be controlled: the notional machine finds a tangible manifestation of its basic properties and capabilities”. Although not studying students working with robots but rather Micro:bits, similar conclusions are forwarded by Cederqvist (2022) who explored the constitution of technological knowledge as students addressed aspects of construction and functionality of a technical solution when programming. Activities that create opportunities for the students to design and adjust different parts of a technical system, such as a programable robot, therefore have the potential to expand students´ understanding of technology.

A relatively well-known programming activity for supporting students´ understanding of input–output-feedback processes and the relation between the code and the physical object is the programming of a robot to follow a line (e.g. Cesaretti et al., 2017; Francis et al., 2016; Merkouris & Konstantinos, 2018; Xia & Zhong, 2018). By using sensors, the robot is programmed to register the surface and depending on the direction of the line, adjust its position and movement by how the wheels rotate. In the study of Gonçalves et al. (2019) the students (age 15–17) who had little prior experiences of programming worked with a task where they were supposed to make a robot follow a curved line. After approximately an hour they were able to solve the task by testing and adjusting the demo code they started from. Although the authors do not discuss in detail the programming strategies the students used, the findings may be in line with Merkouris and Konstantinos (2020) study where it was shown that students with different experiences of programming used different strategies to solve the programming tasks, one of the tasks being to follow a line. More advanced students reused and remixed large parts of code and removed unnecessary blocks while novice students tended to reuse small parts and tested them step by step in cycles. This incremental—iterative approach was not evident among the more advanced students who rather were shown to abstract and modularize different strategies. Both groups were struggling with testing and debugging (Merkouris and Konstantinos, 2020). Some strategies and associated debugging practices may be related to the misconceptions students have when working with programable robots. The students in Komms et al. (2020) study were working with the follow line—problem and two other programming tasks and misconceptions were shown to revolve around recurrent difficulties concerning for example ending the program (understanding the stop() command), multitasking, using delay() commands, and using loops in a proper way. Other difficulties students may encounter may be coupled to the misconception of a one-to-one correspondence between for example the input value of rotating a wheel and the degree measure of a turn (Francis et al., 2016).

In summary, previous research has demonstrated that programmable robots may serve as meaningful and functional tools for teaching and learning technology. Technical systems, being a key component in a technological knowledge, are scrutinized when students debug and readjust their code because of how the parts of the object respond to the program (Cederqvist, 2022). The code itself can also be understood and operationalized as a form of a technical solution, evaluated, and revised due to its functionality (Anderhag et al., 2021). Research has however also shown that students often encounter difficulties when programming feed-back dependent robot movements such as following a line. For example, it has been demonstrated that there are differences in how novice and more experienced students debug their programs. Much of this important knowledge on encountered difficulties and the strategies students apply for overcoming them have been generated from interviews, open-ended questions and pre-, post-test and written representation and the knowledge about how such strategies are constituted in talk and action when students are programming is to our knowledge sparse. The aim of the study is therefore to address this gap on how learning to program can come about in the technology classroom. Our primary interest is how students approach the task of controlling an object through programming and we ask:

There are several possible ways in which the robots in the task could be programmed to follow a curved line. As described above, the students were expected to use the robots' light sensor and they had access to a demo code (Demo code 1, Fig. 2) that they could use as a starting point. In the demo code, the sensor was programmed to continuously register the surface and depending on the colour registered (in School A a black line and white table surface, in School B a green line and a black table surface), the movement was adjusted by the power which the two wheels mounted in parallel rotated. Through the interaction between the sensor and the wheels, the demo code resulted in a zigzag movement of the robot. In this way, the robot moved jerkily forward across the line. The idea of this first strategy, Sensor-Wheels, is thus that algorithms are used to make the two parts, sensor and wheels, work together. It turned out, however, that the students did not use the demo code to any great extent, but instead tinkered with another demo code (Demo code 2, Fig. 2) that they could use as a starting point to make the robot stop when its sensor detected a certain colour. The students that were using Demo code 2 as a starting point were shown to use a second strategy, Sensor-Follow the line. The student groups from school B had access to the same demo codes, but the task differed in that the line did not have the same curved shape (Fig. 1) and the activity of getting the robot to follow a curved line was the whole task for these students. In school A, the activity of following the line was the first in a series of programming tasks. The analysis revealed a third strategy among the student groups in school B, Rotations-Degrees-Wheels. The three strategies are presented below.

At the same time as policymakers argue that knowledge about programming is important in current society (Raptopoulou, 2021), relatively little is known about how programming can be taught by teachers and how it is learnt by students (Anderhag et al., 2021; Vinnervik, 2022a, 2022b). In the Swedish technology subject syllabus, programming is a mean for learning aspects of technology rather than a content and little guidance is provided for teachers regarding what and how programming could be implemented (Vinnervik, 2022a). In this study we have addressed aspects of teaching and learning programming by investigating what strategies students use when they are programming a robot to follow a curved line in the technology class.

Programming an object to follow a line is a rather conventional teaching activity and in accordance with the Swedish syllabus of the technology subject, which states that students should learn to steer an object by using programming (Swedish National Agency for Education, 2018). The study showed that the students used three different strategies for steering the robot: (1) sensor-follow the line, where students discuss and try to implement a code that makes the robot follow the line without communicating with the wheels of the robot, (2) sensor–wheel, where codes were used to create a feedback system in which a light sensor continuously register the surface and where changes in colour resulted in different movements of the wheels, and (3) rotations-degrees-wheel, where students tested and revised the movement along the line by changing the number of rotations of the wheels or how many degrees the wheels should turn. The first strategy could not be realised in a functional program. Our findings support previous studies showing that programmable robots can be usable tools for novice learners (Misirli & Komis, 2014), also students with little or no experiences of programming could participate fruitfully in the activity. The direct feedback on the code by the movement of the robot seemed to be important for student engagement (e.g. Komm et al., 2020) during the activity the students almost exclusively talked about the code in relation to anticipated and observed robot movement. The students thus continuously identified and made judgements on the functionality of the technical solution they worked with, which has been argued to be an important aspect of what constitutes technological knowledge (Anderhag et al., 2021; Björkholm, 2014; Cederqvist, 2022). In line with the findings of Gonçalves et al. (2019), the students in our study spent much time testing and debugging code, our findings contribute by showing how these discussions were aligned with the three strategies. Depending on the strategy, students actively looked for and tested codes affecting different aspects of the sensor-wheel system, such as for example power, rotations or turning. The study did not find the differences between novice and advanced programmers as reported by Merkouris and Konstantinos (2020). In their study, advanced programmers were shown to use and remix large parts of code and while novice students tended to reuse and test small parts of code in iterative cycles. The most salient difference between experienced and novice students in our study was that the advanced students were faster to articulate a reasonable solution to the problem that they worked with, which is in line with the findings of Merkouris and Konstantinos (2020). The more advanced students also spent less time navigating and searching for code blocks, which is expected as they are familiar with the programming environment.

The strategies presented have implications for teaching technology and could be used by teachers when planning and designing learning activities. All student groups succeeded in programming a robot to follow the line and stop at the designated place. However, as the study has showed, “follow” is constituted differently in the three strategies. In strategy 2, follow amounts to a code that adjusts the movement of the robot, so the sensor always is placed above the line. In strategy 3, follow is realized through a code that is fine tuned to a specific route. In this strategy, the independent movement of the robot is not guided by the line but rather the programmer’s evaluation of the movement along the specific line the robot is following. Strategy 1, finally, differs from the other two in that follow is supposed to be realized through an explicit follow code, thus automatically constructing the necessary feedback system between sensor and wheels. Although more studies are needed to see whether the strategies presented here also is evident in other contexts, a starting point for teaching could thus be discussing what follows mean in relation to a self-driving vehicle. The everyday understanding of self-driving vehicles, where following a line or a road means just that, could thus be problematized, and discussed in relation to what kind of feedback system that is necessary for that to happen. It is not surprising that students may look for a code block named “When sensor register black—follow”, and so ignoring that following needs to be continuously orchestrated through the input from the sensor and the output of the wheels. The everyday understanding of programming is likely to support such a position and the code block for stopping the movement of the robot is actually constructed this way in the programming environment used in the study, “When sensor register red—stop”, thus not being transparent of the relation between the sensor and the wheels. Another aspect that teaching could address is the functionality of the code in strategies 2 and 3. Both strategies were successful as they did what they were supposed to do, that is making the robot follow a line. However, as discussed above, the meaning of follow and the code that realizes it can be discussed in relation to different scenarios and settings, such as static and dynamic worlds (e.g. Parson & Sklar, 2004). Strategy 3, for example, only works if the robot always starts at the same place—what are the advantages and disadvantages of such a program? Strategy 2 have the potential of being functional even if the route changes, may there be scenarios when you still want to use another strategy? Would the code work if the line was curved to the left rather than to the right? How could the program—movement of the robot be optimized further? Although not being the scope of the article, it can be noted that the students were occupied with an important aspect of what constitutes technology knowledge, namely how feedback processes operate in a technical system (see Cederqvist, 2022). We can only speculate but some aspects of the strategies here extracted, such as how the students handle the relations between the components of the system or whether they actually make use of feedback processes, may be of relevance also in other contexts than programming.

Finally, the findings of the study are extracted through micro-analysis of student and teacher conversations and other methodologies could provide fruitful avenues for further deepen the understanding on the strategies students use when programming feed-back dependent robot movements. The empirical material of the study comes from two technology classrooms, and we cannot make any claims regarding how prevalent the strategies are. A more quantitative set up could thus be used to explore whether the strategies here presented also are evident in other contexts. Moreover, the fact that the students differed in their previous experiences of programming and also encountered slightly different routes that they should adapt the movements of their robots to, any suggestions to why some students use one strategy before another would be speculations. Further studies are thus needed and an interesting set up would be to explore how we as teachers can plan for activities that better support students in testing and evaluating the functionality of the strategies presented here.