Research projects

Kristian Lindgren – Complex systems group

Department of Energy and Environment, Chalmers

A future energy system that to a larger extent depends on variable renewable energy sources, like wind and solar power, involves new challenges for securing a reliable supply of electricity. The economic basis for such an electricity system differs significantly from the one we have today, and one can expect that prices will be more volatile and risk for shortages in supply increases. The experience from countries where the expansion of solar and wind has been strong exemplifies this. At times with abundant solar electricity supply and low demand, prices have decreased, while at peak demand if supply is low one has seen the opposite (The Economist, 2013; Hirth, 2013). Many owners of fossil fuelled power plants make consider­able losses, and some argue that the electricity market should be restructured in order to allow for investments to secure reliable electricity supply at times of high demand (Newbery, 2010).

In the transition towards a renewable energy system, foresight is needed in order to design policies that facilitate changes that are required for the transition. This also includes improved under­standing of possible consequences of implemented measures and policies—also the unintend­ed ones, like the increased price volatility in the example above. Another example of undesired consequences that has been discussed as a possible effect of climate policies is the food price peaks we have witnessed the passed ten years. The argument has been that the rapidly in­creased demand for bioenergy from agricultural land has led to increased prices on several agricultural commodities, further amplified by subsequent political decisions like export restrictions in certain countries, e.g., India and Vietnam.

The main tool for foresight in the transition of the energy system is numerical modelling. There is in this context a long tradition of optimisation models based on economic cost minimisation or utility maximisation under constraints, like we have done in the Global Energy Transition model GET (Azar et al, 2003, 2006, 2010; Hedenus et al, 2006); for an interactive online demonstration, see Most of the models that are used for assessments of policies are based on rational expectations and perfect knowledge of the future resulting in an equilibrium solution describing the transition. This is part of the equilibrium economics paradigm that in many cases has proven useful, but that in our view is insufficient for a proper analysis of the energy systems transition that is required in order to mitigate climate change. Equilibrium models do not well represent dynamic processes and are thus weak at capturing temporal effects in a system that undergoes a transition.

In a recently started project, we will develop a new set of models in order to be able to capture several dynamic effects in subsystems of high relevance for a transition of the energy system. The two examples men­tioned above, (I) the increased variable supply in the electricity system, and (II) the bioenergy demand effect on the agricultural system, will be investigated in more detail as illustrative examples of our modelling approach, see, e.g., (Bryngelsson and Lindgren, 2013; Jonson et al, 2015; Lundberg et al, 2015; Lindgren et al, 2015).

The new models need to be able to handle short-term variability and mechanisms describing how this affects the economic basis in the system and thus also the investment decisions that determine the development of the system. These details are important in order to capture the effect of the short time scale variability on the long-term development of the whole system. We are thus developing a complement to the traditional energy systems modelling approach that builds on optimisation and equilibrium economics, with models that explicitly includes mechanisms for decisions—mechanisms that do not necessarily assume rational agents but may allow for bounded rationality as a basis for decisions. These models may be formulated on the level of individual agents (citizens, firms, regions, etc), so-called agent-based models, or projections of agents into aggregate variables leading to dynamic simulation models. With an agent-based model, assumptions on rationality and full informaiton may be relaxed, and one may study how non-rational and heterogeneous actors can influence a market (Farmer & Foley, 2009; Arthur, 2014; Page, 2012). One major aim with the project is to critically and carefully investigate the difference between the possible modelling approaches, (i) agent-based modelling, (ii) dynamic aggregate models, and (iii) optimisation or equilibrium modelling. Depending on what questions that are asked, different levels of modelling are appropriate. This will be exemplified with the two mentioned application areas of high relevance for the transformation of the energy system in a climate change perspective. 



1. Arthur, W. B. (2014). Complexity and the Economy. (Oxford University Press).

2. Azar, C., Lindgren, K., Andersson , B. A. (2003). Global energy scenarios meeting stringent CO2 constraints – cost-effective fuel choices in the transportation sector. Energy Policy 31, 961-976.

3. Azar, C., Lindgren, K., Larson, E., and Möllersten, K. (2006). Carbon capture and storage from fossil fuels and biomass — Costs and potential role in stabilizing the atmosphere. Climatic Change 74, 47-79.

4. Azar, C., Lindgren, K., Obersteiner, M., Riah, K., van Vuuren, D.P., den Elzen, K.M.G.J., Möllersten, K., Larson, E. D. (2010). The feasibility of low CO2 concentration targets and the role of bio-energy with carbon capture and storage (BECCS). Climatic Change, 100 (1), 195-202.

5. Bryngelsson, D. K., Lindgren, K. (2013). Why large-scale bioenergy production on marginal land is unfeasible: A conceptual partial equilibrium analysis,” Energy Policy 55, 454-466.

6. Farmer, D. F., Foley, D. (2009). The economy needs agent-based modelling, Nature 460, 685-686. 

7. Hedenus, F., Azar, C., and Lindgren, K. (2006) Induced technological change in a limited foresight optimization model. The Energy Journal 27 special issue 109-122.

8. Hirth, L. (2013). The market value of variable renewables. The effect of solar wind power variability on their relative price. Energy Economics 38, 218–236.

9. Jonson, E., Lundberg, L., Lindgren, K. (2014). Impacts on stability of interdependencies between markets in a cobweb model, paper presented at 10th Artificial Economics Conference, Barcelona, Sept. 1-2, 2014; published in Advances in Artificial Economics, sid 195-205 (2015).

10. Lindgren, K., E. Jonson, and L. Lundberg (2015). Projection of a heterogeneous agent-based production economy model to a closed dynamics of aggregate variables. Advances in Complex Systems.

11. Lundberg, L., Jonson, E., Lindgren, K., Bryngelsson, D., Verendel, V. (2014). A cobweb model of land-use competition between food and bioenergy crops. Journal of Economic Dynamics and Control 53, 1-14 (2015).

12. Newbery, D. (2010). Market design for a large share of wind power. Energy Policy 38, 3131-3134.

13. Page, E. S. (2012). Aggregation in agent-based models of economics. The Knowledge Engi- neering Review 27, 151–162. 

14. The Economist (2013). “How to lose half a trillion euros,” downloadable via

My current research is, to a great extent, centered on autonomous vehicles. This topic involves many challenges, not only the actual path planning and navigation, but also issues related to safety, fuel consumption, efficiency, cooperation between vehicles etc.

In particular, in one of my projects, I'm working on an algorithm for optimizing the movements of a set of dumpers in a mine, such that they will be able to transport material within the mine with optimal efficiency and with complete safety, i.e. without any collisions or incidents. This is a highly dynamic optimization problem since, for example, vehicles can be added to (or removed from) the fleet at any time. Moreover, vehicles can of course also break down or be required to execute an emergency stop, and such events may influence the entire fleet.

More details here:


My group, which consist of myself, Petter Törnberg and Anton Törnberg, works to develop new ways of understanding the evolution of societal systems. We work on several levels – from specific case studies up to basic concept, model and method development. The basis for our work is the realization that innovation is usefully describable on the highly abstract level of “innovation in complex adaptive systems”. In other words there are features of systems under adaptive transformation that appear to be common to just about any instance: animal culture, early hominin culture, modern culture and biological organic evolution.

Darwinism (or more generally “population thinking”) has since long been the basis for such attempts and it is an element also of what we are doing. However we think that Darwinism needs to be scaffolded: evolution – or innovation, whichever one prefers – is not reducible to the dynamics of populations and it does not just emerge out of a microlevel dynamics where the full explanation really resides. The reason for this is really simple and quite concrete: there is no clear scale separation in these systems and so the generated higher levels of organization will have dynamics on timescales that overlap with those on lower levels.

Societal and biological systems have been described in ways that are compatible with our way of thinking for quite some time; not least lately in evolutionary developmental approaches to biology. But these efforts are scattered across very different disciplines and we think we can get a tremendous head start by combining such accounts. So we are working to produce a synthetic theory of innovation with an eye to two different, yet interestingly connected, areas: human evolution and innovation in modern societies. We are also working to develop a community interested in understanding “innovation in complex adaptive systems” more in general.

Using computer simulations in connection with the development, testing and validation of active safety functions is a cost efficient method. However, these simulations must be validated using controlled, physical proving ground tests. In order to make the usage of proving ground tests more efficient, since that is rather expensive and time consuming to use, the NG–TEST project aims to establish a tool chain from computer desktop simulations, via driving simulators and augmented reality testing, to driver-less proving ground tests. It should be possible to seamlessly transfer test scenarios from one tool to another.

A scaled car test track lab, for development and testing of active safety functions for cars and trucks, is currently under development in the Adaptive systems research group at Chalmers University of Technology. The scaled car test track environment is intended as a tool for, for example, rapid verification of scenarios before executing them in the real automated vehicles on the full–scale proving ground.

Within the NG–TEST context, the STT lab constitutes an efficient rapid prototyping tool for design and verification of scenarios, especially in those situations where driver behavior aspects are important. Furthermore, it is also a powerful visualization tool, which could benefit the developers, sponsors and customers of active safety functions. A large part of the research that will be carried out using the STT lab will focus on how such a lab can be used in connection with active safety system research, and how it could benefit the NG–TEST process. Compared with a full–scale vehicle, the down-scaling certainly give raise to non-linear scaling of the behaviors of the model car, e.g. maximum cornering speeds will exceed those of full– scale cars. Therefore, the research will mainly focus on issues related to test scenario and driver behavior aspects, rather than questions where very realistic vehicle dynamics is important.

The basic principles for processing a raw material has always been controlling chemical mixing, temperature cycling, etc. The function of an end product is then created by cutting, forging and imposing a form on the material at the macroscopic level. This two-step paradigm is about to change. Our recently acquired capability to synthesize nanoscale particles with almost arbitrary shape and interactions has opened up for self-assembly of complex structures and novel metamaterials. For the first time in history we have direct control over the building blocks that form the material itself and determine its characteristics. However, we are still far from the fine-tuned mechanisms for self-organization designed by evolution. Most work on self-assembly is experimental or simulation-based, but to reach further and design self-assembling systems rivaling those in nature, we cannot rely on trial and error or rules of thumb. In our projects we therefore focus on theoretical methods for understanding and controlling self-assembly. For example, we have shown that both prediction and design is possible in one-component systems with isotropic interactions. We, among other things, devised a method that allowed us to find a design for the first self-assembly of a chiral lattice (crystal structure breaking mirror symmetry) from rotationally symmetric particles.

We are also working on particles with anisotropic interactions in the form of patchy colloids, micrometer sized particles covered with interacting patches. We have both developed a method for predicting patch formation on colloids and studied the use of patchy colloids as building blocks for self-assembly.

In this project, we are developing tools to help the elderly and the visually impaired. We have developed a so-called partner robot (see the figure) that can function as a conversation partner. One can, for example, ask the robot to read the news, whereupon it collects information from the Internet and then reads it aloud, while interacting with the user (thus, the user can, at any time, change subject, stop the robot from reading, skip certain paragraphs etc.).


Importantly, one can use this robot without any knowledge of computers or robotics; essentially, one speaks to the robot in the same way as one would to a person. The robot can interpret human speech (albeit with a slightly limited vocabulary), and it also has an adjustable synthetic voice. In addition to the robot, we have recently begun developing a so called intelligent agent, in the form of an animated face on a computer screen, with the same aim in mind, namely that it should function as a conversation partner for the elderly and the visually impaired.